# Changelog

All notable changes to this project will be documented in this file.

This project adheres to [Semantic Versioning](https://semver.org/spec/v2.0.0.html).

## [0.24.1] - 2026-09-16

### 🐛 Bug Fixes

- *(benchmark)* Correct Locust concurrency measurements and sweeps ([#2323](https://github.com/NVIDIA-NeMo/Guardrails/issues/2323))
- *(content-safety)* Parse Nemotron 3.5 responses ([#2370](https://github.com/NVIDIA-NeMo/Guardrails/issues/2370))
- *(server)* Return overload response for IORails admission shedding ([#2334](https://github.com/NVIDIA-NeMo/Guardrails/issues/2334))
- *(rails)* Load config.py for combined configurations ([#2371](https://github.com/NVIDIA-NeMo/Guardrails/issues/2371))


## [0.24.0] - 2026-08-25

### 🚀 Features

- *(library)* Add F5 Guardrails integration ([#2105](https://github.com/NVIDIA-NeMo/Guardrails/issues/2105))
- *(server)* Add `/v1/health` and `/healthz` health-check endpoints ([#2169](https://github.com/NVIDIA-NeMo/Guardrails/issues/2169))
- *(actions)* Add the engine-neutral `RailOutcome` allow, block, and transform contract ([#2150](https://github.com/NVIDIA-NeMo/Guardrails/issues/2150))
- *(iorails)* Add message checks ([#2059](https://github.com/NVIDIA-NeMo/Guardrails/issues/2059))
- *(iorails)* Support model-level `default_headers` and  `default_query` fields for inference requests ([#2296](https://github.com/NVIDIA-NeMo/Guardrails/issues/2296), [#2220](https://github.com/NVIDIA-NeMo/Guardrails/issues/2220))
- *(self-check)* Run multiple self-check rails with per-rail namespaced task prompts ([#1874](https://github.com/NVIDIA-NeMo/Guardrails/issues/1874), [#2175](https://github.com/NVIDIA-NeMo/Guardrails/issues/2175))
- *(rails)* Add the typed rail manifest contract ([#2157](https://github.com/NVIDIA-NeMo/Guardrails/issues/2157))
- *(iorails)* [**breaking**] Return `GenerationResponse` when `GenerationOptions` are provided for non-streaming inference. Pass message lists with `messages=` instead of positionally ([#2178](https://github.com/NVIDIA-NeMo/Guardrails/issues/2178))
- *(iorails)* Add support for all actions without Colang runtime dependency ([#2241](https://github.com/NVIDIA-NeMo/Guardrails/issues/2241), [#2246](https://github.com/NVIDIA-NeMo/Guardrails/issues/2246), [#2253](https://github.com/NVIDIA-NeMo/Guardrails/issues/2253), [#2261](https://github.com/NVIDIA-NeMo/Guardrails/issues/2261), [#2264](https://github.com/NVIDIA-NeMo/Guardrails/issues/2264), [#2288](https://github.com/NVIDIA-NeMo/Guardrails/issues/2288))
- *(http)* Add the canonical outbound HTTP client and request lifecycle ([#2209](https://github.com/NVIDIA-NeMo/Guardrails/issues/2209), [#2210](https://github.com/NVIDIA-NeMo/Guardrails/issues/2210))
- *(llm)* Add shared model telemetry and an instrumented model decorator ([#2214](https://github.com/NVIDIA-NeMo/Guardrails/issues/2214))
- *(http)* Add outbound client tracing and metrics ([#2219](https://github.com/NVIDIA-NeMo/Guardrails/issues/2219))
- *(server)* Add output-rail checking mode to `/v1/checks` ([#2205](https://github.com/NVIDIA-NeMo/Guardrails/issues/2205))

### 🐛 Bug Fixes

- *(colang)* Raise `ValueError` for an unbalanced closing bracket in `split_args` ([#2145](https://github.com/NVIDIA-NeMo/Guardrails/issues/2145))
- *(streaming)* Fail closed when rail actions fail ([#2152](https://github.com/NVIDIA-NeMo/Guardrails/issues/2152))
- *(library)* [**breaking**] Make bundled manifest flows portable across Colang versions; custom Colang 1 flows using legacy space-separated Cleanlab, Fiddler, or GCP action names must switch to snake_case names ([#2185](https://github.com/NVIDIA-NeMo/Guardrails/issues/2185))
- *(actions)* Make `RailDecision` JSON serializable ([#2194](https://github.com/NVIDIA-NeMo/Guardrails/issues/2194))
- *(iorails)* Preserve normalized usage and provider metadata in `include_metadata=True` streams ([#2198](https://github.com/NVIDIA-NeMo/Guardrails/issues/2198))
- *(server)* [**breaking**] Remove inline configuration from `/v1/checks`; select a server-loaded configuration with `config_id` or use the server default ([#2228](https://github.com/NVIDIA-NeMo/Guardrails/issues/2228))
- *(server)* Return OpenAI-compatible HTTP error envelopes while preserving provider status, code, parameter, and retry metadata ([#1832](https://github.com/NVIDIA-NeMo/Guardrails/issues/1832))
- *(server)* Return HTTP 400 for thread IDs without a datastore and preserve the configured main model API-key field during request model injection ([#2240](https://github.com/NVIDIA-NeMo/Guardrails/issues/2240))
- *(server)* Honor request-level stop parameters and reject chat messages that omit a role before dispatch ([#2266](https://github.com/NVIDIA-NeMo/Guardrails/issues/2266))
- *(hf-classifier)* Preserve HTTP retry instrumentation ([#2291](https://github.com/NVIDIA-NeMo/Guardrails/issues/2291))
- *(llm)* Preserve streaming usage on terminal chunks ([#2295](https://github.com/NVIDIA-NeMo/Guardrails/issues/2295))
- *(content-safety)* Surface response parsing errors instead of silently allowing malformed results ([#2294](https://github.com/NVIDIA-NeMo/Guardrails/issues/2294))
- *(checks)* Reject unsatisfiable `rail_types` with HTTP 422 instead of silently passing ([#2276](https://github.com/NVIDIA-NeMo/Guardrails/issues/2276))
- *(server)* Preserve configured main model fields when a request specifies a model ([#2298](https://github.com/NVIDIA-NeMo/Guardrails/issues/2298))
- *(telemetry)* Derive built-in feature reporting from rail manifests ([#2301](https://github.com/NVIDIA-NeMo/Guardrails/issues/2301))
- *(iorails)* Map provider, timeout, connection, response, and streaming failures to safe client errors while preserving HTTP status and retry metadata ([#2306](https://github.com/NVIDIA-NeMo/Guardrails/issues/2306))
- *(logging)* Leave application logging unchanged unless Guardrails verbose logging is explicitly enabled ([#2310](https://github.com/NVIDIA-NeMo/Guardrails/issues/2310))
- *(benchmark)* Let the Locust CLI target the Guardrails server ([#2309](https://github.com/NVIDIA-NeMo/Guardrails/issues/2309))
- *(server)* [**breaking**] Validate Chat Completions messages by role and reject internal event payloads, unexpected fields, and unsupported audio requests; send only supported OpenAI-compatible message shapes ([#2311](https://github.com/NVIDIA-NeMo/Guardrails/issues/2311))
- *(server)* Inline `<think>` tags when reasoning content is present ([#2316](https://github.com/NVIDIA-NeMo/Guardrails/issues/2316))
- *(iorails)* Handle reasoning-only non-streaming responses ([#2317](https://github.com/NVIDIA-NeMo/Guardrails/issues/2317))
- *(iorails)* Distinguish policy blocks from rail execution failures so callers can identify retryable provider outages ([#2318](https://github.com/NVIDIA-NeMo/Guardrails/issues/2318))

### 💼 Other

- *(dependencies)* [**breaking**] Remove the `hf-classifier` install extra and its packages from `all`; install `transformers` and `torch` directly for the local classifier backend ([#2137](https://github.com/NVIDIA-NeMo/Guardrails/issues/2137))

### 🚜 Refactor

- *(actions)* [**breaking**] Migrate built-in rail actions and streaming bypasses to `RailOutcome` and remove `@action(output_mapping=...)`; custom rail actions using `output_mapping` must return explicit `RailOutcome` decisions ([#2151](https://github.com/NVIDIA-NeMo/Guardrails/issues/2151))
- *(library)* Migrate built-in rail configuration declarations to manifests without changing the public configuration shape ([#2181](https://github.com/NVIDIA-NeMo/Guardrails/issues/2181))
- *(self-check)* Name the rail selector variant consistently from flow to action ([#2182](https://github.com/NVIDIA-NeMo/Guardrails/issues/2182))
- *(library)* Make moderation outcomes flow-independent ([#2183](https://github.com/NVIDIA-NeMo/Guardrails/issues/2183))
- *(library)* Make context-bloat verdicts flow-independent ([#2190](https://github.com/NVIDIA-NeMo/Guardrails/issues/2190))
- *(actions)* Resolve manifest actions lazily ([#2186](https://github.com/NVIDIA-NeMo/Guardrails/issues/2186))
- *(library)* Decouple Llama Guard and Patronus Lynx model injection ([#2236](https://github.com/NVIDIA-NeMo/Guardrails/issues/2236))
- *(http)* Migrate built-in integration request helpers and vendor actions to the canonical managed HTTP client lifecycle ([#2211](https://github.com/NVIDIA-NeMo/Guardrails/issues/2211), [#2212](https://github.com/NVIDIA-NeMo/Guardrails/issues/2212))
- *(iorails)* Migrate IORails to manifest-compiled `RailOutcome` actions with shared model and HTTP dependencies ([#2286](https://github.com/NVIDIA-NeMo/Guardrails/issues/2286))
- *(rails)* Make shared action dependencies explicit ([#2293](https://github.com/NVIDIA-NeMo/Guardrails/issues/2293))

### 📚 Documentation

- *(installation)* Remove obsolete Annoy C++ compiler guidance ([#2148](https://github.com/NVIDIA-NeMo/Guardrails/issues/2148))
- *(guardrail-catalog)* Document running multiple self-check rails ([#2176](https://github.com/NVIDIA-NeMo/Guardrails/issues/2176))
- Document the out-of-scope security policy ([#2192](https://github.com/NVIDIA-NeMo/Guardrails/issues/2192))
- Document the `hf_classifier` rail and its local and remote backends ([#1969](https://github.com/NVIDIA-NeMo/Guardrails/issues/1969))
- Clarify repository agent guidance ([#2218](https://github.com/NVIDIA-NeMo/Guardrails/issues/2218))
- Publish a community telemetry snapshot ([#2302](https://github.com/NVIDIA-NeMo/Guardrails/issues/2302))
- Add Fern documentation versioning ([#2287](https://github.com/NVIDIA-NeMo/Guardrails/issues/2287))
- *(benchmark)* Add a Locust mock-server quickstart ([#2307](https://github.com/NVIDIA-NeMo/Guardrails/issues/2307))
- *(actions)* Document `RailOutcome`-based rail actions ([#2258](https://github.com/NVIDIA-NeMo/Guardrails/issues/2258))
- *(rails)* Document rail manifests and action-backed surfaces ([#2259](https://github.com/NVIDIA-NeMo/Guardrails/issues/2259))
- *(http)* Document canonical outbound HTTP clients ([#2260](https://github.com/NVIDIA-NeMo/Guardrails/issues/2260))
- *(iorails)* Document `default_headers` and `default_query` ([#2322](https://github.com/NVIDIA-NeMo/Guardrails/issues/2322))
- *(iorails)* Document `GenerationOptions` and `GenerationResponse` support ([#2328](https://github.com/NVIDIA-NeMo/Guardrails/issues/2328))
- *(iorails)* Document message checks ([#2329](https://github.com/NVIDIA-NeMo/Guardrails/issues/2329))
- *(iorails)* Document per-surface LLMRails and IORails compatibility, fallback conditions, and configuration-dependent rail support ([#2330](https://github.com/NVIDIA-NeMo/Guardrails/issues/2330))

### 🧪 Testing

- Add unit coverage for eval, evaluate, and actions ([#2078](https://github.com/NVIDIA-NeMo/Guardrails/issues/2078))
- *(langchain)* Make provider-drift snapshots version-series-aware ([#2149](https://github.com/NVIDIA-NeMo/Guardrails/issues/2149))
- *(ci)* Require fix pull requests to demonstrate a failing-first regression test ([#2153](https://github.com/NVIDIA-NeMo/Guardrails/issues/2153))
- *(ci)* Enforce test coverage quality gates ([#2161](https://github.com/NVIDIA-NeMo/Guardrails/issues/2161))
- *(recorded)* Add runtime checks and inclusion tests ([#2167](https://github.com/NVIDIA-NeMo/Guardrails/issues/2167))
- *(ci)* Separate pull-request coverage checks from the test matrix ([#2180](https://github.com/NVIDIA-NeMo/Guardrails/issues/2180))
- *(recorded)* Require cassettes to be sanitizer fixed points ([#2184](https://github.com/NVIDIA-NeMo/Guardrails/issues/2184))
- *(langchain)* Record the Meta partner chat provider in the drift snapshot ([#2191](https://github.com/NVIDIA-NeMo/Guardrails/issues/2191))
- Improve test isolation ([#2204](https://github.com/NVIDIA-NeMo/Guardrails/issues/2204))
- *(rails)* Enforce built-in manifest conformance ([#2189](https://github.com/NVIDIA-NeMo/Guardrails/issues/2189))
- *(server)* Add OpenAI API specification conformance tracking ([#2197](https://github.com/NVIDIA-NeMo/Guardrails/issues/2197))
- *(frameworks)* Prevent the default framework from leaking between tests ([#2215](https://github.com/NVIDIA-NeMo/Guardrails/issues/2215))
- *(llm)* Move call tests to module mirrors ([#2265](https://github.com/NVIDIA-NeMo/Guardrails/issues/2265))

## [0.23.0] - 2026-07-01

### 🚀 Features

- *(library)* Add lightweight Hugging Face classifier rails for input, output, and retrieval, with local Transformers, vLLM, KServe, and FMS backends ([#1853](https://github.com/NVIDIA-NeMo/Guardrails/issues/1853))
- *(embeddings)* Replace Annoy with exact NumPy search and add migration benchmarks ([#1957](https://github.com/NVIDIA-NeMo/Guardrails/issues/1957), [#1958](https://github.com/NVIDIA-NeMo/Guardrails/issues/1958))
- *(iorails)* Add opt-in OpenTelemetry content capture and LLM request, response, and usage attributes ([#1972](https://github.com/NVIDIA-NeMo/Guardrails/issues/1972), [#2009](https://github.com/NVIDIA-NeMo/Guardrails/issues/2009))
- *(iorails)* Add streaming and non-streaming tool calling and local rails for validating tool calls and results ([#2016](https://github.com/NVIDIA-NeMo/Guardrails/issues/2016), [#2024](https://github.com/NVIDIA-NeMo/Guardrails/issues/2024), [#2030](https://github.com/NVIDIA-NeMo/Guardrails/issues/2030), [#2058](https://github.com/NVIDIA-NeMo/Guardrails/issues/2058))
- *(library)* Add context bloat detection rail ([#1941](https://github.com/NVIDIA-NeMo/Guardrails/issues/1941))
- *(server)* Add `/v1/checks` endpoint for standalone input and output rail validation ([#2013](https://github.com/NVIDIA-NeMo/Guardrails/issues/2013))
- *(server)* Add tool calling support ([#1942](https://github.com/NVIDIA-NeMo/Guardrails/issues/1942))
- *(examples)* Introduce NIM-based example notebooks, retire superseded ones ([#1906](https://github.com/NVIDIA-NeMo/Guardrails/issues/1906))
- *(library)* Add Polygraf PII detection and masking integration ([#1693](https://github.com/NVIDIA-NeMo/Guardrails/issues/1693))

### 🐛 Bug Fixes

- *(library)* Fix regex detection during output streaming so matches block correctly without raising `TypeError` ([#1932](https://github.com/NVIDIA-NeMo/Guardrails/issues/1932), [#1937](https://github.com/NVIDIA-NeMo/Guardrails/issues/1937))
- *(actions)* Avoid empty-string crash in create_event ([#1701](https://github.com/NVIDIA-NeMo/Guardrails/issues/1701))
- *(iorails)* Make OTEL recording best-effort ([#1997](https://github.com/NVIDIA-NeMo/Guardrails/issues/1997))
- *(docs)* Skip Fern bash-script tests on Windows ([#2017](https://github.com/NVIDIA-NeMo/Guardrails/issues/2017))
- *(iorails)* Apply inference-time llm_params on top of Model.parameters in ModelEngine ([#2020](https://github.com/NVIDIA-NeMo/Guardrails/issues/2020))
- *(llm)* Handle multi-line bot say responses in flow continuation([#1650](https://github.com/NVIDIA-NeMo/Guardrails/issues/1650))
- *(generation)* Use correct task enum for stop tokens in generate_value ([#1699](https://github.com/NVIDIA-NeMo/Guardrails/issues/1699))
- *(colang)* Guard ' or' line continuation at end of file ([#1947](https://github.com/NVIDIA-NeMo/Guardrails/issues/1947))
- *(iorails)* Add no-op events_history_cache when IORails is used ([#2072](https://github.com/NVIDIA-NeMo/Guardrails/issues/2072))
- *(llmrails)* Load library files deterministically ([#1975](https://github.com/NVIDIA-NeMo/Guardrails/issues/1975))
- *(embeddings)* EmbeddingsCache.from_dict drops store_config on round-trip ([#1951](https://github.com/NVIDIA-NeMo/Guardrails/issues/1951))
- *(eval)* Use safe dumper and yaml load ([#2082](https://github.com/NVIDIA-NeMo/Guardrails/issues/2082))
- *(streaming)* Pass user content to output rails ([#2081](https://github.com/NVIDIA-NeMo/Guardrails/issues/2081))
- *(streaming)* Avoid duplicate usage metadata chunk ([#2079](https://github.com/NVIDIA-NeMo/Guardrails/issues/2079))
- *(streaming)* Don't reuse resolved action parameters across output-rail chunks or requests ([#1935](https://github.com/NVIDIA-NeMo/Guardrails/issues/1935), [#1943](https://github.com/NVIDIA-NeMo/Guardrails/issues/1943))
- *(ci)* Update README version during releases ([#2104](https://github.com/NVIDIA-NeMo/Guardrails/issues/2104))
- *(llmrails)* Preserve tool calls for LLMRails tool rails ([#2073](https://github.com/NVIDIA-NeMo/Guardrails/issues/2073))
- *(langchain)* OpenAI Responses API and Harmony response format support ([#2102](https://github.com/NVIDIA-NeMo/Guardrails/issues/2102))

### 💼 Other

- Stop bundling examples and repo files in the wheel (10x smaller) ([#2069](https://github.com/NVIDIA-NeMo/Guardrails/issues/2069))
- Exclude repository agent instruction files from source and wheel packages ([#2111](https://github.com/NVIDIA-NeMo/Guardrails/issues/2111))

### 🚜 Refactor

- Refine the Guardrails public API and deprecate direct access to internal `LLMRails` attributes ([#1933](https://github.com/NVIDIA-NeMo/Guardrails/issues/1933))
- [**breaking**] Require Pydantic `>=2.5,<3.0` and migrate validators and model APIs to Pydantic 2 ([#967](https://github.com/NVIDIA-NeMo/Guardrails/issues/967))

### 📚 Documentation

- Clarify NGC_API_KEY handling for local GLiNER/PII NIM deployment ([#1945](https://github.com/NVIDIA-NeMo/Guardrails/issues/1945))
- Migrate documentation to Fern, document the publishing workflow, and complete link and template cleanup ([#1973](https://github.com/NVIDIA-NeMo/Guardrails/issues/1973), [#2015](https://github.com/NVIDIA-NeMo/Guardrails/issues/2015), [#2018](https://github.com/NVIDIA-NeMo/Guardrails/issues/2018), [#2019](https://github.com/NVIDIA-NeMo/Guardrails/issues/2019))
- *(readme)* Fix broken links to the Guardrails website ([#2046](https://github.com/NVIDIA-NeMo/Guardrails/issues/2046))
- *(skills)* Add skills ([#2025](https://github.com/NVIDIA-NeMo/Guardrails/issues/2025))
- *(iorails)* Telemetry - Span Reference Docs ([#2098](https://github.com/NVIDIA-NeMo/Guardrails/issues/2098))
- *(iorails)* Tool calling docs ([#2099](https://github.com/NVIDIA-NeMo/Guardrails/issues/2099))
- *(iorails)* Telemetry - Content Capture docs ([#2083](https://github.com/NVIDIA-NeMo/Guardrails/issues/2083))

### 🧪 Testing

- Make xdist the default Makefile test path ([#1970](https://github.com/NVIDIA-NeMo/Guardrails/issues/1970))
- Isolate flaky wall-clock perf tests behind a perf marker ([#2070](https://github.com/NVIDIA-NeMo/Guardrails/issues/2070))
- *(recorded)* Add a replay harness, client cassette coverage, public API coverage, and rails library coverage ([#1974](https://github.com/NVIDIA-NeMo/Guardrails/issues/1974), [#1976](https://github.com/NVIDIA-NeMo/Guardrails/issues/1976), [#1977](https://github.com/NVIDIA-NeMo/Guardrails/issues/1977), [#1978](https://github.com/NVIDIA-NeMo/Guardrails/issues/1978))
- *(langchain)* Make provider compat tests version-tolerant, add drift canary ([#2071](https://github.com/NVIDIA-NeMo/Guardrails/issues/2071))
- Support aiohttp 3.14 in aioresponses mocks ([#2091](https://github.com/NVIDIA-NeMo/Guardrails/issues/2091))
- Remove flaky streaming timing diagnostic ([#2097](https://github.com/NVIDIA-NeMo/Guardrails/issues/2097))

## [0.22.0] - 2026-05-22

### 🚀 Features

- *(iorails)* IORails support for streaming output rails ([#1765](https://github.com/NVIDIA-NeMo/Guardrails/issues/1765), [#1766](https://github.com/NVIDIA-NeMo/Guardrails/issues/1766))
- *(iorails)* IORails OpenTelemetry tracing support ([#1793](https://github.com/NVIDIA-NeMo/Guardrails/issues/1793), [#1794](https://github.com/NVIDIA-NeMo/Guardrails/issues/1794), [#1798](https://github.com/NVIDIA-NeMo/Guardrails/issues/1798))
- *(iorails)* IORails OpenTelemetry token-level metrics support ([#1812](https://github.com/NVIDIA-NeMo/Guardrails/issues/1812), [#1846](https://github.com/NVIDIA-NeMo/Guardrails/issues/1846))
- *(iorails)* IORails reasoning model support ([#1842](https://github.com/NVIDIA-NeMo/Guardrails/issues/1842), [#1843](https://github.com/NVIDIA-NeMo/Guardrails/issues/1843))
- *(llm)* Add LangChain adapter and framework registry ([#1759](https://github.com/NVIDIA-NeMo/Guardrails/issues/1759))
- *(llm)* Add streaming tool call accumulation and LLMResponse parity ([#1789](https://github.com/NVIDIA-NeMo/Guardrails/issues/1789))
- *(llm)* Add default framework with OpenAI-compatible client ([#1797](https://github.com/NVIDIA-NeMo/Guardrails/issues/1797))
- *(llm/frameworks)* Validate framework on registration ([#1863](https://github.com/NVIDIA-NeMo/Guardrails/issues/1863))
- *(types)* Add framework-agnostic LLM type system ([#1745](https://github.com/NVIDIA-NeMo/Guardrails/issues/1745))
- *(compat)* Transitional compat layer to migrate from 0.21 to 0.22+ ([#1841](https://github.com/NVIDIA-NeMo/Guardrails/issues/1841))
- *(testing)* Add public testing surface under nemoguardrails.testing ([#1860](https://github.com/NVIDIA-NeMo/Guardrails/issues/1860))
- *(api)* Canonical top-level imports for LLM types and registry functions ([#1882](https://github.com/NVIDIA-NeMo/Guardrails/issues/1882))
- *(config)* Forbade extra fields in GLiNER rails configs ([#1898](https://github.com/NVIDIA-NeMo/Guardrails/issues/1898))
- *(framework)* Support Azure as a first-class default framework preset ([#1896](https://github.com/NVIDIA-NeMo/Guardrails/issues/1896))

### 🐛 Bug Fixes

- [**breaking**] Reject Colang 2.0 public runtime state ([#1885](https://github.com/NVIDIA-NeMo/Guardrails/issues/1885))
- *(server)* Prioritize env var API key over forwarded client header ([#1688](https://github.com/NVIDIA-NeMo/Guardrails/issues/1688))
- *(utils)* Removing extra space from UtteranceBotActionScriptUpdated ([#1708](https://github.com/NVIDIA-NeMo/Guardrails/issues/1708))
- *(actions)* Remove redundant embedding search in generate_user_intent ([#1754](https://github.com/NVIDIA-NeMo/Guardrails/issues/1754))
- *(llmrails)* Backfill embedding model params into search provider config (fixes stale KB cache) ([#1753](https://github.com/NVIDIA-NeMo/Guardrails/issues/1753))
- *(embeddings)* Persist in-memory embedding cache instance across calls ([#1755](https://github.com/NVIDIA-NeMo/Guardrails/issues/1755))
- *(ci)* Pin baseline x86-64 compiler target to prevent SIGILL on cached venvs ([#1785](https://github.com/NVIDIA-NeMo/Guardrails/issues/1785))
- *(tests)* Use asyncio.run instead of get_event_loop in middleware tests ([#1804](https://github.com/NVIDIA-NeMo/Guardrails/issues/1804))
- *(actions)* Guard bot message extraction against composite specs ([#1810](https://github.com/NVIDIA-NeMo/Guardrails/issues/1810))
- *(llm)* Drop stop param for OpenAI reasoning models ([#1811](https://github.com/NVIDIA-NeMo/Guardrails/issues/1811))
- *(llmrails)* Scope no main LLM warning to generation path ([#1813](https://github.com/NVIDIA-NeMo/Guardrails/issues/1813))
- *(chat-ui)* Replace Chatbot UI with Chainlit ([#1734](https://github.com/NVIDIA-NeMo/Guardrails/issues/1734))
- *(library)* Unblock reasoning models in self-check and content-safety actions ([#1816](https://github.com/NVIDIA-NeMo/Guardrails/issues/1816))
- *(llm)* Drop temperature/stop and rename max_tokens for OpenAI reasoning models ([#1837](https://github.com/NVIDIA-NeMo/Guardrails/issues/1837))
- *(build)* Remove chat-ui references from wheel build script ([#1835](https://github.com/NVIDIA-NeMo/Guardrails/issues/1835))
- *(ci)* Override annoy's -march=native to actually enforce baseline x86-64 ([#1839](https://github.com/NVIDIA-NeMo/Guardrails/issues/1839))
- *(llm/clients)* Retry on stale event loop binding ([#1840](https://github.com/NVIDIA-NeMo/Guardrails/issues/1840))
- *(llm/frameworks)* Point users to LangChain when DefaultFramework has no base_url ([#1865](https://github.com/NVIDIA-NeMo/Guardrails/issues/1865))
- *(taskmanager)* Preserve multimodal list content in vision safety prompts ([#1815](https://github.com/NVIDIA-NeMo/Guardrails/issues/1815))
- *(actions)* Extract text from multimodal events in colang history ([#1636](https://github.com/NVIDIA-NeMo/Guardrails/issues/1636))
- *(iorails)* Route to LLMRails if Guardrails inits with provided LLM ([#1844](https://github.com/NVIDIA-NeMo/Guardrails/issues/1844))
- *(iorails)* Strip trailing /v1 from base_url to avoid doubled path ([#1862](https://github.com/NVIDIA-NeMo/Guardrails/issues/1862))
- *(iorails)* Add all LLMRails methods, `llm` and `runtime` getters to Guardrails facade ([#1886](https://github.com/NVIDIA-NeMo/Guardrails/issues/1886), [#1889](https://github.com/NVIDIA-NeMo/Guardrails/issues/1889))
- *(iorails)* Annotate cancelled OTEL spans with error ([#1897](https://github.com/NVIDIA-NeMo/Guardrails/issues/1897))

### 🚜 Refactor

- *(llm)* [**breaking**] Atomic switch to LLMModel protocol ([#1760](https://github.com/NVIDIA-NeMo/Guardrails/issues/1760))
- *(iorails)* [**breaking**] Move AsyncWorkQueue from Guardrails to IORails ([#1817](https://github.com/NVIDIA-NeMo/Guardrails/issues/1817))
- *(deps)* [**breaking**] Demote LangChain and LangChain-providers from core to dev ([#1806](https://github.com/NVIDIA-NeMo/Guardrails/issues/1806))
- *(iorails)* Return LLMResponse(Chunk) from ModelEngine ([#1827](https://github.com/NVIDIA-NeMo/Guardrails/issues/1827))
- *(iorails)* Refactor ModelManager ([#1778](https://github.com/NVIDIA-NeMo/Guardrails/issues/1778))
- *(iorails)* Refactor Guardrails and IORails for top-level import and clean separation ([#1893](https://github.com/NVIDIA-NeMo/Guardrails/issues/1893))
- *(iorails)* Refactor RailsManager and Nemoguard Actions ([#1762](https://github.com/NVIDIA-NeMo/Guardrails/issues/1762))
- *(llm)* Rename generate/stream to generate_async/stream_async ([#1769](https://github.com/NVIDIA-NeMo/Guardrails/issues/1769))
- *(llm)* Remove LangChain imports from core modules ([#1770](https://github.com/NVIDIA-NeMo/Guardrails/issues/1770))
- *(llm)* Move LangChain implementations into integrations/langchain/ ([#1772](https://github.com/NVIDIA-NeMo/Guardrails/issues/1772))
- *(llm)* Framework-owned provider registry ([#1773](https://github.com/NVIDIA-NeMo/Guardrails/issues/1773))
- *(llm)* Share OpenAI reasoning-model classifier across adapters ([#1836](https://github.com/NVIDIA-NeMo/Guardrails/issues/1836))
- *(llm)* Reorganize llm package into clients/models/frameworks ([#1801](https://github.com/NVIDIA-NeMo/Guardrails/issues/1801))
- *(llm/clients)* Return HTTPResponse(body, headers, status_code) from_apost ([#1830](https://github.com/NVIDIA-NeMo/Guardrails/issues/1830))
- *(llm/default_framework)* Split reset() into aclose() + clear_providers() ([#1829](https://github.com/NVIDIA-NeMo/Guardrails/issues/1829))
- *(tests)* Framework-agnostic test infrastructure ([#1790](https://github.com/NVIDIA-NeMo/Guardrails/issues/1790))
- *(deps)* Make server-only dependencies optional ([#1689](https://github.com/NVIDIA-NeMo/Guardrails/issues/1689))
- *(jailbreak)* Use onnx instead of pickle to load model ([#1715](https://github.com/NVIDIA-NeMo/Guardrails/issues/1715))
- *(logging)* Remove LangChain LoggingCallbackHandler dependency ([#1616](https://github.com/NVIDIA-NeMo/Guardrails/issues/1616))

### 📚 Documentation

- Document release notes for 0.21 and additional details ([#1726](https://github.com/NVIDIA-NeMo/Guardrails/issues/1726))
- *(middleware)* Fix incorrect example query and expected output in agent-middleware guide ([#1784](https://github.com/NVIDIA-NeMo/Guardrails/issues/1784))
- *(iorails)* OTEL Logging page ([#1807](https://github.com/NVIDIA-NeMo/Guardrails/issues/1807))
- Fix jira 407 ([#1809](https://github.com/NVIDIA-NeMo/Guardrails/issues/1809))
- Update README ([#1820](https://github.com/NVIDIA-NeMo/Guardrails/issues/1820))
- Mark LangChain integration as opt-in in 0.22 entry-point docs ([#1856](https://github.com/NVIDIA-NeMo/Guardrails/issues/1856))
- Documentation for langchain decoupling ([#1854](https://github.com/NVIDIA-NeMo/Guardrails/issues/1854))
- *(configure-rails)* Align with 0.22 DefaultFramework / LangChain split ([#1855](https://github.com/NVIDIA-NeMo/Guardrails/issues/1855))
- *(custom-initialization)* Add customLLM and customFramework guides ([#1857](https://github.com/NVIDIA-NeMo/Guardrails/issues/1857))
- *(examples)* Align example configs and deployment docs with 0.22 DefaultFramework / LangChain split ([#1858](https://github.com/NVIDIA-NeMo/Guardrails/issues/1858))
- *(iorails)* IORails OTEL Metrics ([#1864](https://github.com/NVIDIA-NeMo/Guardrails/issues/1864))
- *(iorails)* Speculative Generation ([#1876](https://github.com/NVIDIA-NeMo/Guardrails/issues/1876))
- *(migration)* Explain the "No default base_url" config-load error ([#1881](https://github.com/NVIDIA-NeMo/Guardrails/issues/1881))
- *(iorails)* Use Guardrails entry-point not IORails ([#1892](https://github.com/NVIDIA-NeMo/Guardrails/issues/1892))
- Prometheus client install instructions ([#1894](https://github.com/NVIDIA-NeMo/Guardrails/issues/1894))
- *(guardrails)* Document max_tokens fallback and reasoning model guidance ([#1833](https://github.com/NVIDIA-NeMo/Guardrails/issues/1833))
- *(colang-1)* Fix Hello World tutorial issues (NGUARD-666) ([#1834](https://github.com/NVIDIA-NeMo/Guardrails/issues/1834))
- *(telemetry)* Document anonymous usage reporting ([#1891](https://github.com/NVIDIA-NeMo/Guardrails/issues/1891))
- Add prompts.yml to code snippets ([#1904](https://github.com/NVIDIA-NeMo/Guardrails/issues/1904))
- Update Benchmark README with updated configs ([#1905](https://github.com/NVIDIA-NeMo/Guardrails/issues/1905))

### 🧪 Testing

- *(llm)* Probe OpenAI API to validate _is_openai_reasoning_model ([#1814](https://github.com/NVIDIA-NeMo/Guardrails/issues/1814))
- *(llm)* Expand reasoning-model param probe + regenerate baseline ([#1838](https://github.com/NVIDIA-NeMo/Guardrails/issues/1838))
- *(telemetry)* Add smoke driver and fixtures  ([#1879](https://github.com/NVIDIA-NeMo/Guardrails/issues/1879))

### ⚙️ Miscellaneous Tasks

- Restore original 2023-2026 copyright dates on moved files ([#1831](https://github.com/NVIDIA-NeMo/Guardrails/issues/1831))
- Include scripts in docker image ([#1902](https://github.com/NVIDIA-NeMo/Guardrails/issues/1902))

## [0.21.0] - 2026-03-12

### 🚀 Features

- *(library)* Update Trend Micro Vision One AI Guard official endpoint ([#1546](https://github.com/NVIDIA-NeMo/Guardrails/issues/1546))
- *(llmrails)* Add check_async method for input/output rails validation ([#1605](https://github.com/NVIDIA-NeMo/Guardrails/issues/1605))
- *(server)* Make guardrails server OpenAI compatible ([#1340](https://github.com/NVIDIA-NeMo/Guardrails/issues/1340))
- New top-level scaffold ([#1613](https://github.com/NVIDIA-NeMo/Guardrails/issues/1613))
- Add Async work queue ([#1620](https://github.com/NVIDIA-NeMo/Guardrails/issues/1620))
- *(integration)* Add GuardrailsMiddleware for LangChain agent ([#1606](https://github.com/NVIDIA-NeMo/Guardrails/issues/1606))
- *(library)* Update Fiddler Guardrails API to match new specification ([#1619](https://github.com/NVIDIA-NeMo/Guardrails/issues/1619))
- *(library)* Add CrowdStrike AIDR community integration ([#1601](https://github.com/NVIDIA-NeMo/Guardrails/issues/1601))
- *(iorails)* Introduce IORails optimized Input/Output rail engine. Supports non-streaming parallel nemoguard input/output rails (content-safety, topic-safety, jailbreak detection) ([#1638](https://github.com/NVIDIA-NeMo/Guardrails/issues/1638), [#1649](https://github.com/NVIDIA-NeMo/Guardrails/issues/1649), [#1654](https://github.com/NVIDIA-NeMo/Guardrails/issues/1654), [#1656](https://github.com/NVIDIA-NeMo/Guardrails/issues/1656), [#1658](https://github.com/NVIDIA-NeMo/Guardrails/issues/1658), [#1660](https://github.com/NVIDIA-NeMo/Guardrails/issues/1660), [#1661](https://github.com/NVIDIA-NeMo/Guardrails/issues/1661), [#1674](https://github.com/NVIDIA-NeMo/Guardrails/issues/1674))
- *(server)* Add OpenAI compatible v1/models endpoint ([#1637](https://github.com/NVIDIA-NeMo/Guardrails/issues/1637))
- *(benchmark)* Add Locust stress-test ([#1629](https://github.com/NVIDIA-NeMo/Guardrails/issues/1629))
- *(jailbreak)* Validate Jailbreak Detection config at create-time ([#1675](https://github.com/NVIDIA-NeMo/Guardrails/issues/1675))
- *(library)* Add PolicyAI Integration for Content Moderation ([#1576](https://github.com/NVIDIA-NeMo/Guardrails/issues/1576))

### 🐛 Bug Fixes

- *(server)* Make openai an optional server-only dependency ([#1623](https://github.com/NVIDIA-NeMo/Guardrails/issues/1623))
- *(actions)* Rename generate_next_step to generate_next_steps for task-specific LLM support ([#1603](https://github.com/NVIDIA-NeMo/Guardrails/issues/1603))
- *(library)* Add `valid` alias to action results in GuardrailsAI integration ([#1578](https://github.com/NVIDIA-NeMo/Guardrails/issues/1578)) ([#1611](https://github.com/NVIDIA-NeMo/Guardrails/issues/1611))
- *(llm)* Filter stop parameter for OpenAI reasoning models ([#1653](https://github.com/NVIDIA-NeMo/Guardrails/issues/1653))
- *(logging)* Show cache hits in Stats log and fix duplicate metadata restore ([#1666](https://github.com/NVIDIA-NeMo/Guardrails/issues/1666))
- *(cache)* Make cache stats log visible in verbose mode ([#1667](https://github.com/NVIDIA-NeMo/Guardrails/issues/1667))
- *(library)* Use bot refuse to respond in gliner PII detection flows ([#1671](https://github.com/NVIDIA-NeMo/Guardrails/issues/1671))
- *(streaming)* Handle None stop tokens in streaming handler ([#1685](https://github.com/NVIDIA-NeMo/Guardrails/issues/1685))
- *(streaming)* Handle dict chunks in RollingBuffer.format_chunks ([#1687](https://github.com/NVIDIA-NeMo/Guardrails/issues/1687))
- *(middleware)* Handle MODIFIED status in GuardrailsMiddleware instead of silently dropping it ([#1714](https://github.com/NVIDIA-NeMo/Guardrails/issues/1714))

### 🚜 Refactor

- *(streaming)* Remove LangChain callback dependencies from StreamingHandler ([#1547](https://github.com/NVIDIA-NeMo/Guardrails/issues/1547))
- *(streaming)* Remove ChatNVIDIA streaming patch ([#1607](https://github.com/NVIDIA-NeMo/Guardrails/issues/1607))
- *(streaming)* [**breaking**] Remove stream_usage and fix streaming metadata capture ([#1624](https://github.com/NVIDIA-NeMo/Guardrails/issues/1624))

### ⚡ Performance

- *(actions)* Lazy initialization of embedding indexes ([#1572](https://github.com/NVIDIA-NeMo/Guardrails/issues/1572))

### ⚙️ Miscellaneous Tasks

- Update Pangea User-Agent repo URL ([#1595](https://github.com/NVIDIA-NeMo/Guardrails/issues/1595)) ([#1610](https://github.com/NVIDIA-NeMo/Guardrails/issues/1610))
- *(jailbreak)* Update dependencies for jailbreak detection docker container. ([#1596](https://github.com/NVIDIA-NeMo/Guardrails/issues/1596))
- Remove multi_kb example ([#1673](https://github.com/NVIDIA-NeMo/Guardrails/issues/1673))
- *(iorails)* Increase work queue concurrency and depth ([#1674](https://github.com/NVIDIA-NeMo/Guardrails/issues/1674))
- *(docs)* Remove AI Virtual Assistant Blueprint notebook ([#1682](https://github.com/NVIDIA-NeMo/Guardrails/issues/1682))
- Update dependencies ahead of v0.21 release ([#1617](https://github.com/NVIDIA-NeMo/Guardrails/issues/1617))

## [0.20.0] - 2026-01-22

### 🚀 Features

- *(llm)* Propagate model and base URL in LLMCallException; improve error handling ([#1502](https://github.com/NVIDIA-NeMo/Guardrails/issues/1502))
- *(content_safety)* Add support to auto select multilingual refusal bot messages ([#1530](https://github.com/NVIDIA-NeMo/Guardrails/issues/1530))
- *(library)* Adding GLiNER for PII detection (open alternative to PrivateAI) ([#1545](https://github.com/NVIDIA-NeMo/Guardrails/issues/1545))
- *(benchmark)* Implement Mock LLM streaming ([#1564](https://github.com/NVIDIA-NeMo/Guardrails/issues/1564))
- *(library)* Add reasoning guardrail connector ([#1565](https://github.com/NVIDIA-NeMo/Guardrails/issues/1565))

### 🐛 Bug Fixes

- *(models)* Surface relevant exception when initializing langchain model ([#1516](https://github.com/NVIDIA-NeMo/Guardrails/issues/1516))
- *(llm)* Filter temperature parameter for OpenAI reasoning models ([#1526](https://github.com/NVIDIA-NeMo/Guardrails/issues/1526))
- *(bot-thinking)* Tackle bug with reasoning trace leak across llm calls ([#1582](https://github.com/NVIDIA-NeMo/Guardrails/issues/1582))
- *(providers)* Handle langchain 1.2.1 dict type for _SUPPORTED_PROVIDERS ([#1589](https://github.com/NVIDIA-NeMo/Guardrails/issues/1589))

### 🚜 Refactor

- *(streaming)* [**breaking**] Drop streaming field from config ([#1538](https://github.com/NVIDIA-NeMo/Guardrails/issues/1538))

### ⚙️ Miscellaneous Tasks

- *(test)* Reduce default pytest log level from DEBUG to WARNING ([#1523](https://github.com/NVIDIA-NeMo/Guardrails/issues/1523))
- *(docker)* Upgrade to Python 3.12-slim base image ([#1522](https://github.com/NVIDIA-NeMo/Guardrails/issues/1522))
- Run pre-commits to update license date for 2026 ([#1562](https://github.com/NVIDIA-NeMo/Guardrails/issues/1562))
- Move Benchmark code to top-level ([#1559](https://github.com/NVIDIA-NeMo/Guardrails/issues/1559))
- Update repo to <https://github.com/NVIDIA-NeMo/Guardrails> ([#1594](https://github.com/NVIDIA-NeMo/Guardrails/issues/1594))

## [0.19.0] - 2025-12-03

### 🚀 Features

- Support langchain v1 ([#1472](https://github.com/NVIDIA-NeMo/Guardrails/issues/1472))
- *(llm)* Add LangChain 1.x content blocks support for reasoning and tool calls ([#1496](https://github.com/NVIDIA-NeMo/Guardrails/issues/1496))
- *(benchmark)* Add Procfile to run Guardrails and mock LLMs ([#1490](https://github.com/NVIDIA-NeMo/Guardrails/issues/1490))
- *(benchmark)*: Add AIPerf run script (([#1501](https://github.com/NVIDIA-NeMo/Guardrails/issues/1501)))

### 🐛 Bug Fixes

- *(llm)* Add async streaming support to ChatNVIDIA provider patch ([#1504](https://github.com/NVIDIA-NeMo/Guardrails/issues/1504))
- ensure stream_async background task completes before exit ([#1508](https://github.com/NVIDIA-NeMo/Guardrails/issues/1508))
- *(cli)* Fix TypeError in v2.x chat due to incorrect State/dict conversion ([#1509](https://github.com/NVIDIA-NeMo/Guardrails/issues/1509))
- *(llmrails)*: skip output rails when dialog disabled and no bot_message provided ([#1518](https://github.com/NVIDIA-NeMo/Guardrails/issues/1518))
- *(llm)*: ensure that stop token is not ignored if llm_params is None ([#1529](https://github.com/NVIDIA-NeMo/Guardrails/issues/1529))

### ⚙️ Miscellaneous Tasks

- *(llm)* Remove deprecated llm_params module ([#1475](https://github.com/NVIDIA-NeMo/Guardrails/issues/1475))

### ◀️ Revert

- *(llm)* Remove custom HTTP headers patch now in langchain-nvidia-ai-endpoints v0.3.19 ([#1503](https://github.com/NVIDIA-NeMo/Guardrails/issues/1503))

## [0.18.0] - 2025-11-06

### 🚀 Features

- *(bot-thinking)* Implement BotThinking events to process reasoning traces in Guardrails ([#1431](https://github.com/NVIDIA-NeMo/Guardrails/issues/1431)), ([#1432](https://github.com/NVIDIA-NeMo/Guardrails/issues/1432)), ([#1434](https://github.com/NVIDIA-NeMo/Guardrails/issues/1434)).
- *(embeddings)* Add Azure OpenAI embedding provider ([#702](https://github.com/NVIDIA-NeMo/Guardrails/issues/702)).
- *(embeddings)* Add Cohere embedding integration ([#1305](https://github.com/NVIDIA-NeMo/Guardrails/issues/1305)).
- *(embeddings)* Add Google embedding integration ([#1304](https://github.com/NVIDIA-NeMo/Guardrails/issues/1304)).
- *(library)* Add Cisco AI Defense integration ([#1433](https://github.com/NVIDIA-NeMo/Guardrails/issues/1433)).
- *(cache)* Add in-memory LFU caches for content-safety, topic-control, and jailbreak detection models ([#1436](https://github.com/NVIDIA-NeMo/Guardrails/issues/1436)), ([#1456](https://github.com/NVIDIA-NeMo/Guardrails/issues/1456)),  ([#1457](https://github.com/NVIDIA-NeMo/Guardrails/issues/1457)), ([#1458](https://github.com/NVIDIA-NeMo/Guardrails/issues/1458)).
- *(llm)* Add automatic provider inference for LangChain LLMs ([#1460](https://github.com/NVIDIA-NeMo/Guardrails/issues/1460)).
- *(llm)* Add custom HTTP headers support to ChatNVIDIA provider ([#1461](https://github.com/NVIDIA-NeMo/Guardrails/issues/1461)).

### 🐛 Bug Fixes

- *(config)* Validate content safety and topic control configs at creation time ([#1450](https://github.com/NVIDIA-NeMo/Guardrails/issues/1450)).
- *(jailbreak)* Capitalization of `Snowflake` in use of `snowflake-arctic-embed-m-long` name. ([#1464](https://github.com/NVIDIA-NeMo/Guardrails/issues/1464)).
- *(runtime)* Ensure stop flag is set for policy violations in parallel rails ([#1467](https://github.com/NVIDIA-NeMo/Guardrails/issues/1467)).
- *(llm)* [**breaking**] Extract reasoning traces to separate field instead of prepending ([#1468](https://github.com/NVIDIA-NeMo/Guardrails/issues/1468)).
- *(streaming)* [**breaking**] Raise error when stream_async used with disabled output rails streaming ([#1470](https://github.com/NVIDIA-NeMo/Guardrails/issues/1470)).
- *(llm)* Add fallback extraction for reasoning traces from <think> tags ([#1474](https://github.com/NVIDIA-NeMo/Guardrails/issues/1474)).
- *(runtime)* Set stop flag for exception-based rails in parallel mode ([#1487](https://github.com/NVIDIA-NeMo/Guardrails/issues/1487)).

### 🚜 Refactor

- [**breaking**] Replace reasoning trace extraction with LangChain additional_kwargs ([#1427](https://github.com/NVIDIA-NeMo/Guardrails/issues/1427))

### 📚 Documentation

- *(examples)* Add Nemoguard in-memory cache configuration example ([#1459](https://github.com/NVIDIA-NeMo/Guardrails/issues/1459)), ([#1480](https://github.com/NVIDIA-NeMo/Guardrails/issues/1480)).
- Add guide for bot reasoning guardrails ([#1479](https://github.com/NVIDIA-NeMo/Guardrails/issues/1479)).
- Update LLM reasoning traces configuration ([#1483](https://github.com/NVIDIA-NeMo/Guardrails/issues/1483)).

### 🧪 Testing

- Add mock embedding provider tests ([#1446](https://github.com/NVIDIA-NeMo/Guardrails/issues/1446))
- *(cli)* Add comprehensive CLI test suite and reorganize files ([#1339](https://github.com/NVIDIA-NeMo/Guardrails/issues/1339))
- Skip FastEmbed tests when not in live mode ([#1462](https://github.com/NVIDIA-NeMo/Guardrails/issues/1462))
- Fix flaky stats logging interval timing test ([#1463](https://github.com/NVIDIA-NeMo/Guardrails/issues/1463))
- Restore test that was skipped due to Colang 2.0 serialization issue ([#1449](https://github.com/NVIDIA-NeMo/Guardrails/issues/1449))

### ⚙️ Miscellaneous Tasks

- Resolve PyPI publish workflow trigger and reliability issues ([#1443](https://github.com/NVIDIA-NeMo/Guardrails/issues/1443))
- Fix sparse checkout for publish pypi workflow ([#1444](https://github.com/NVIDIA-NeMo/Guardrails/issues/1444))
- Drop Python 3.9 support ahead of October 2025 EOL ([#1426](https://github.com/NVIDIA-NeMo/Guardrails/issues/1426))
- *(types)* Add type-annotations and pre-commit checks for tracing ([#1388](https://github.com/NVIDIA-NeMo/Guardrails/issues/1388)), logging ([#1395](https://github.com/NVIDIA-NeMo/Guardrails/issues/1395)), kb  ([#1385](https://github.com/NVIDIA-NeMo/Guardrails/issues/1385)), cli ([#1380](https://github.com/NVIDIA-NeMo/Guardrails/issues/1380)), embeddings ([#1383](https://github.com/NVIDIA-NeMo/Guardrails/issues/1383)), server ([#1397](https://github.com/NVIDIA-NeMo/Guardrails/issues/1397)), and llm ([#1394](https://github.com/NVIDIA-NeMo/Guardrails/issues/1394)) code.
- Update insert licenser pe-commit-hooks to use current year ([#1452](https://github.com/NVIDIA-NeMo/Guardrails/issues/1452)).
- *(library)* Remove unused vllm requirements.txt files ([#1466](https://github.com/NVIDIA-NeMo/Guardrails/issues/1466)).

## [0.17.0] - 2025-10-09

### 🚀 Features

- *(tool-calling)* Add tool call passthrough support in LLMRails ([#1364](https://github.com/NVIDIA-NeMo/Guardrails/issues/1364))
- *(runnable-rails)* Complete rewrite of RunnableRails with full LangChain Runnable protocol support ([#1366](https://github.com/NVIDIA-NeMo/Guardrails/issues/1366), [#1369](https://github.com/NVIDIA-NeMo/Guardrails/issues/1369), [#1370](https://github.com/NVIDIA-NeMo/Guardrails/issues/1370), [#1405](https://github.com/NVIDIA-NeMo/Guardrails/issues/1405))
- *(tool-rails)* Add support for tool output rails and validation ([#1382](https://github.com/NVIDIA-NeMo/Guardrails/issues/1382))
- *(tool-rails)* Implement tool input rails for tool message validation and processing ([#1386](https://github.com/NVIDIA-NeMo/Guardrails/issues/1386))
- *(library)* Add Trend Micro Vision One AI Application Security community integration ([#1355](https://github.com/NVIDIA-NeMo/Guardrails/issues/1355))
- *(llm)* Pass llm params directly ([#1387](https://github.com/NVIDIA-NeMo/Guardrails/issues/1387))

### 🐛 Bug Fixes

- *(jailbreak)* Handle URL joining with/without trailing slashes ([#1346](https://github.com/NVIDIA-NeMo/Guardrails/issues/1346))
- *(logging)* Handle missing id and task in verbose logs ([#1343](https://github.com/NVIDIA-NeMo/Guardrails/issues/1343))
- *(library)* Fix import package declaration to new cleanlab-tlm name ([#1401](https://github.com/NVIDIA-NeMo/Guardrails/issues/1401))
- *(logging)* Add "Tool" type to message sender labeling ([#1412](https://github.com/NVIDIA-NeMo/Guardrails/issues/1412))
- *(logging)* Correct message type formatting in logs ([#1416](https://github.com/NVIDIA-NeMo/Guardrails/issues/1416))

### 🚜 Refactor

- *(llm)* Remove LLMs isolation for actions ([#1408](https://github.com/NVIDIA-NeMo/Guardrails/issues/1408))

### 📚 Documentation

- *(examples)* Add NeMoGuard safety rails config example for Colang 1.0 ([#1365](https://github.com/NVIDIA-NeMo/Guardrails/issues/1365))
- Add hardware reqs ([#1411](https://github.com/NVIDIA-NeMo/Guardrails/issues/1411))
- Add tools integration guide ([#1414](https://github.com/NVIDIA-NeMo/Guardrails/issues/1414))
- *(langgraph)* Add integration guide for LangGraph ([#1422](https://github.com/NVIDIA-NeMo/Guardrails/issues/1422))
- *(langchain)* Update with full support and add tool calling guide … ([#1419](https://github.com/NVIDIA-NeMo/Guardrails/issues/1419))
- *(langgraph)* Clarify tool examples and replace calculate_math with multiply ([#1439](https://github.com/NVIDIA-NeMo/Guardrails/issues/1439))

### ⚙️ Miscellaneous Tasks

- *(docs)* Update v0.16.0 release date in changelog ([#1377](https://github.com/NVIDIA-NeMo/Guardrails/issues/1377))
- *(docs)* Add link to demo.py script in Getting-Started section ([#1399](https://github.com/NVIDIA-NeMo/Guardrails/issues/1399))
- *(types)* Type-clean rails (86 errors) ([#1396](https://github.com/NVIDIA-NeMo/Guardrails/issues/1396))
- *(jailbreak-detection)* Update transformers and torch ([#1417](https://github.com/NVIDIA-NeMo/Guardrails/issues/1417))
- *(types)* Type-clean /actions (189 errors) ([#1361](https://github.com/NVIDIA-NeMo/Guardrails/issues/1361))
- *(docs)* Update repository owner ([#1425](https://github.com/NVIDIA-NeMo/Guardrails/issues/1425))

## [0.16.0] - 2025-09-05

### 🚀 Features

- *(llmrails)* Support method chaining by returning self from LLMRails.register_* methods ([#1296](https://github.com/NVIDIA-NeMo/Guardrails/issues/1296))
- Add Pangea AI Guard community integration ([#1300](https://github.com/NVIDIA-NeMo/Guardrails/issues/1300))
- *(llmrails)* Isolate LLMs only for configured actions ([#1342](https://github.com/NVIDIA-NeMo/Guardrails/issues/1342))
- Enhance tracing system with OpenTelemetry semantic conventions ([#1331](https://github.com/NVIDIA-NeMo/Guardrails/issues/1331))
- Add GuardrailsAI community integration ([#1298](https://github.com/NVIDIA-NeMo/Guardrails/issues/1298))

### 🐛 Bug Fixes

- *(models)* Suppress langchain_nvidia_ai_endpoints warnings ([#1371](https://github.com/NVIDIA-NeMo/Guardrails/issues/1371))
- *(tracing)* Respect the user-provided log options regardless of tracing configuration
- *(config)* Ensure adding RailsConfig objects handles None values ([#1328](https://github.com/NVIDIA-NeMo/Guardrails/issues/1328))
- *(config)* Add handling for config directory with `.yml`/`.yaml` extension ([#1293](https://github.com/NVIDIA-NeMo/Guardrails/issues/1293))
- *(colang)* Apply guardrails transformations to LLM inputs and bot outputs. ([#1297](https://github.com/NVIDIA-NeMo/Guardrails/issues/1297))
- *(topic_safety)* Handle InternalEvent objects in topic safety actions for Colang 2.0 ([#1335](https://github.com/NVIDIA-NeMo/Guardrails/issues/1335))
- *(prompts)* Prevent IndexError when LLM provided via constructor with empty models config ([#1334](https://github.com/NVIDIA-NeMo/Guardrails/issues/1334))
- *(llmrails)* Handle LLM models without model_kwargs field in isolation ([#1336](https://github.com/NVIDIA-NeMo/Guardrails/issues/1336))
- *(llmrails)* Move LLM isolation setup to after KB initialization ([#1348](https://github.com/NVIDIA-NeMo/Guardrails/issues/1348))

### 🚜 Refactor

- *(llm)* Move get_action_details_from_flow_id from llmrails.py to utils.py ([#1341](https://github.com/NVIDIA-NeMo/Guardrails/issues/1341))

### 📚 Documentation

- Integrate with multilingual NIM ([#1354](https://github.com/NVIDIA-NeMo/Guardrails/issues/1354))
- *(tracing)* Update tracing notebooks with VDR feedback ([#1376](https://github.com/NVIDIA-NeMo/Guardrails/issues/1376))
- Add kv cache reuse documentation ([#1330](https://github.com/NVIDIA-NeMo/Guardrails/issues/1330))
- *(examples)* Add Colang 2.0 example for sensitive data detection ([#1301](https://github.com/NVIDIA-NeMo/Guardrails/issues/1301))
- Add extra slash to jailbreak detect nim_base_url([#1345](https://github.com/NVIDIA-NeMo/Guardrails/issues/1345))
- Add tracing notebook ([#1337](https://github.com/NVIDIA-NeMo/Guardrails/issues/1337))
- Jaeger tracing notebook ([#1353](https://github.com/NVIDIA-NeMo/Guardrails/issues/1353))
- *(examples)* Add NeMoGuard rails config for colang 2 ([#1289](https://github.com/NVIDIA-NeMo/Guardrails/issues/1289))
- *(tracing)* Add OpenTelemetry span format guide ([#1350](https://github.com/NVIDIA-NeMo/Guardrails/issues/1350))
- Add GuardrailsAI integration user guide and example ([#1357](https://github.com/NVIDIA-NeMo/Guardrails/issues/1357))

### 🧪 Testing

- *(jailbreak)* Add missing pytest.mark.asyncio decorators ([#1352](https://github.com/NVIDIA-NeMo/Guardrails/issues/1352))

### ⚙️ Miscellaneous Tasks

- *(docs)* Rename test_csl.py to csl.py ([#1347](https://github.com/NVIDIA-NeMo/Guardrails/issues/1347))

## [0.15.0] - 2025-08-08

### 🚀 Features

- *(tracing)* [**breaking**] Update tracing to use otel api ([#1269](https://github.com/NVIDIA-NeMo/Guardrails/issues/1269))
- *(streaming)* Implement parallel streaming output rails execution ([#1263](https://github.com/NVIDIA-NeMo/Guardrails/issues/1263), [#1324](https://github.com/NVIDIA-NeMo/Guardrails/pull/1324))
- *(streaming)* Support external async token generators ([#1286](https://github.com/NVIDIA-NeMo/Guardrails/issues/1286))
- Support parallel rails execution ([#1234](https://github.com/NVIDIA-NeMo/Guardrails/issues/1234), [#1323](https://github.com/NVIDIA-NeMo/Guardrails/pull/1323))

### 🐛 Bug Fixes

- *(streaming)* Resolve word concatenation in streaming output rails ([#1259](https://github.com/NVIDIA-NeMo/Guardrails/issues/1259))
- *(streaming)* Enable token usage tracking for streaming LLM calls ([#1264](https://github.com/NVIDIA-NeMo/Guardrails/issues/1264), [#1285](https://github.com/NVIDIA-NeMo/Guardrails/issues/1285))
- *(tracing)* Prevent mutation of user options when tracing is enabled ([#1273](https://github.com/NVIDIA-NeMo/Guardrails/issues/1273))
- *(rails)* Prevent LLM parameter contamination in rails ([#1306](https://github.com/NVIDIA-NeMo/Guardrails/issues/1306))

### 📚 Documentation

- Release notes 0.14.1 ([#1272](https://github.com/NVIDIA-NeMo/Guardrails/issues/1272))
- Update guardrails-library.md to include Clavata as a third party API ([#1294](https://github.com/NVIDIA-NeMo/Guardrails/issues/1294))
- *(streaming)* Add section on token usage tracking ([#1282](https://github.com/NVIDIA-NeMo/Guardrails/issues/1282))
- Add parallel rail section and split config page ([#1295](https://github.com/NVIDIA-NeMo/Guardrails/issues/1295))
- Show complete prompts.yml content in getting started tutorial ([#1311](https://github.com/NVIDIA-NeMo/Guardrails/issues/1311))
- *(tracing)* Update and streamline tracing guide ([#1307](https://github.com/NVIDIA-NeMo/Guardrails/issues/1307))

### ⚙️ Miscellaneous Tasks

- *(dependabot)* Remove dependabot configuration ([#1281](https://github.com/NVIDIA-NeMo/Guardrails/issues/1281))
- *(CI)* Add release workflow ([#1309](https://github.com/NVIDIA-NeMo/Guardrails/issues/1309), [#1318](https://github.com/NVIDIA-NeMo/Guardrails/issues/1318))

## [0.14.1] - 2025-07-02

### 🚀 Features

- *(jailbreak)* Add direct API key configuration support ([#1260](https://github.com/NVIDIA-NeMo/Guardrails/issues/1260))

### 🐛 Bug Fixes

- *(jailbreak)* Lazy load jailbreak detection dependencies ([#1223](https://github.com/NVIDIA-NeMo/Guardrails/issues/1223),)
- *(llmrails)* Constructor LLM should not skip loading other config models ([#1221](https://github.com/NVIDIA-NeMo/Guardrails/issues/1221), [#1247](https://github.com/NVIDIA-NeMo/Guardrails/issues/1247), [#1250](https://github.com/NVIDIA-NeMo/Guardrails/issues/1250), [#1258](https://github.com/NVIDIA-NeMo/Guardrails/issues/1258))
- *(content_safety)* Replace try-except with iterable unpacking for policy violations ([#1207](https://github.com/NVIDIA-NeMo/Guardrails/issues/1207))
- *(jailbreak)* Pin numpy==1.23.5 for scikit-learn compatibility ([#1249](https://github.com/NVIDIA-NeMo/Guardrails/issues/1249))
- *(output_parsers)* Iterable unpacking compatibility in content safety parsers ([#1242](https://github.com/NVIDIA-NeMo/Guardrails/issues/1242))

### 📚 Documentation

- More heading levels so RNs resolve links ([#1228](https://github.com/NVIDIA-NeMo/Guardrails/issues/1228))
- Update docs version ([#1219](https://github.com/NVIDIA-NeMo/Guardrails/issues/1219))
- Fix jailbreak detection build instructions ([#1248](https://github.com/NVIDIA-NeMo/Guardrails/issues/1248))
- Change ABC bot link at docs ([#1261](https://github.com/NVIDIA-NeMo/Guardrails/issues/1261))

### 🧪 Testing

- Fix async test failures in cache embeddings and buffer strategy tests ([#1237](https://github.com/NVIDIA-NeMo/Guardrails/issues/1237))
- *(content_safety)* Add tests for content safety actions ([#1240](https://github.com/NVIDIA-NeMo/Guardrails/issues/1240))

### ⚙️ Miscellaneous Tasks

- Update pre-commit-hooks to v5.0.0 ([#1238](https://github.com/NVIDIA-NeMo/Guardrails/issues/1238))

## [0.14.0] - 2025-05-28

### 🚀 Features

- Change topic following prompt to allow chitchat ([#1097](https://github.com/NVIDIA-NeMo/Guardrails/issues/1097))
- Validate model name configuration ([#1084](https://github.com/NVIDIA-NeMo/Guardrails/issues/1084))
- Add support for langchain partner and community chat models ([#1085](https://github.com/NVIDIA-NeMo/Guardrails/issues/1085))
- Add fuzzy find provider capability to cli ([#1088](https://github.com/NVIDIA-NeMo/Guardrails/issues/1088))
- Add code injection detection to guardrails library ([#1091](https://github.com/NVIDIA-NeMo/Guardrails/issues/1091))
- Add clavata community integration ([#1027](https://github.com/NVIDIA-NeMo/Guardrails/issues/1027))
- Implement validation to forbid dialog rails with reasoning traces ([#1137](https://github.com/NVIDIA-NeMo/Guardrails/issues/1137))
- Load yara lazily to avoid action dispatcher error ([#1162](https://github.com/NVIDIA-NeMo/Guardrails/issues/1162))
- Add support for system messages to RunnableRails ([#1106](https://github.com/NVIDIA-NeMo/Guardrails/issues/1106))
- Add api_key_env_var to Model, pass in kwargs to langchain initializer ([#1142](https://github.com/NVIDIA-NeMo/Guardrails/issues/1142))
- Add inline YARA rules support ([#1164](https://github.com/NVIDIA-NeMo/Guardrails/issues/1164))
- [**breaking**] Add support for preserving and optionally applying guardrails to reasoning traces ([#1145](https://github.com/NVIDIA-NeMo/Guardrails/issues/1145))
- Prevent reasoning traces from contaminating LLM prompt history ([#1169](https://github.com/NVIDIA-NeMo/Guardrails/issues/1169))
- Add RailException support to injection detection and improve error handling ([#1178](https://github.com/NVIDIA-NeMo/Guardrails/issues/1178))
- Add Nemotron model support with message-based prompts ([#1199](https://github.com/NVIDIA-NeMo/Guardrails/issues/1199))

### 🐛 Bug Fixes

- Correct task name for self_check_facts ([#1040](https://github.com/NVIDIA-NeMo/Guardrails/issues/1040))
- Error in LLMRails with tracing enabled ([#1103](https://github.com/NVIDIA-NeMo/Guardrails/issues/1103))
- Self check output colang 1 flow ([#1126](https://github.com/NVIDIA-NeMo/Guardrails/issues/1126))
- Use ValueError in TaskPrompt to resolve TypeError raised by Pydantic ([#1132](https://github.com/NVIDIA-NeMo/Guardrails/issues/1132))
- Correct dialog rails activation logic ([#1161](https://github.com/NVIDIA-NeMo/Guardrails/issues/1161))
- Allow reasoning traces when embeddings_only is True ([#1170](https://github.com/NVIDIA-NeMo/Guardrails/issues/1170))
- Prevent explain_info overwrite during stream_async ([#1194](https://github.com/NVIDIA-NeMo/Guardrails/issues/1194))
- Colang 2 issues in community integrations ([#1140](https://github.com/NVIDIA-NeMo/Guardrails/issues/1140))
- Ensure proper asyncio task cleanup in test_streaming_handler.py ([#1182](https://github.com/NVIDIA-NeMo/Guardrails/issues/1182))

### 🚜 Refactor

- Reorganize HuggingFace provider structure ([#1083](https://github.com/NVIDIA-NeMo/Guardrails/issues/1083))
- Remove support for deprecated nemollm engine ([#1076](https://github.com/NVIDIA-NeMo/Guardrails/issues/1076))
- [**breaking**] Remove deprecated return_context argument ([#1147](https://github.com/NVIDIA-NeMo/Guardrails/issues/1147))
- Rename `remove_thinking_traces` field to `remove_reasoning_traces` ([#1176](https://github.com/NVIDIA-NeMo/Guardrails/issues/1176))
- Update deprecated field handling  for remove_thinking_traces ([#1196](https://github.com/NVIDIA-NeMo/Guardrails/issues/1196))
- Introduce END_OF_STREAM sentinel and update handling ([#1185](https://github.com/NVIDIA-NeMo/Guardrails/issues/1185))

### 📚 Documentation

- Remove markup from code block ([#1081](https://github.com/NVIDIA-NeMo/Guardrails/issues/1081))
- Replace img tag with Markdown images ([#1087](https://github.com/NVIDIA-NeMo/Guardrails/issues/1087))
- Remove NeMo Service (nemollm) documentation ([#1077](https://github.com/NVIDIA-NeMo/Guardrails/issues/1077))
- Update cleanlab integration description ([#1080](https://github.com/NVIDIA-NeMo/Guardrails/issues/1080))
- Add providers fuzzy search cli command ([#1089](https://github.com/NVIDIA-NeMo/Guardrails/issues/1089))
- Clarify purpose of model parameters field in configuration guide ([#1181](https://github.com/NVIDIA-NeMo/Guardrails/issues/1181))
- Output rails are supported with streaming ([#1007](https://github.com/NVIDIA-NeMo/Guardrails/issues/1007))
- Add mention of Nemotron ([#1200](https://github.com/NVIDIA-NeMo/Guardrails/issues/1200))
- Fix output rail doc ([#1159](https://github.com/NVIDIA-NeMo/Guardrails/issues/1159))
- Revise GS example in getting started doc ([#1146](https://github.com/NVIDIA-NeMo/Guardrails/issues/1146))
- Possible update to injection detection ([#1144](https://github.com/NVIDIA-NeMo/Guardrails/issues/1144))

### ⚙️ Miscellaneous Tasks

- Dynamically set version using importlib.metadata ([#1072](https://github.com/NVIDIA-NeMo/Guardrails/issues/1072))
- Add link to topic control config and prompts ([#1098](https://github.com/NVIDIA-NeMo/Guardrails/issues/1098))
- Reorganize GitHub workflows for better test coverage ([#1079](https://github.com/NVIDIA-NeMo/Guardrails/issues/1079))
- Add summary jobs for workflow branch protection ([#1120](https://github.com/NVIDIA-NeMo/Guardrails/issues/1120))
- Add Adobe Analytics configuration ([#1138](https://github.com/NVIDIA-NeMo/Guardrails/issues/1138))
- Fix and revert poetry lock to its stable state ([#1133](https://github.com/NVIDIA-NeMo/Guardrails/issues/1133))
- Add Codecov integration to workflows ([#1143](https://github.com/NVIDIA-NeMo/Guardrails/issues/1143))
- Add Python 3.12 and 3.13 test jobs to gitlab workflow ([#1171](https://github.com/NVIDIA-NeMo/Guardrails/issues/1171))
- Identify OS packages to install in contribution guide([#1136](https://github.com/NVIDIA-NeMo/Guardrails/issues/1136))
- Remove Got It AI from ToC in 3rd party docs([#1213](https://github.com/NVIDIA-NeMo/Guardrails/issues/1213))

## [0.13.0] - 2025-03-25

### 🚀 Features

- Support models with reasoning traces ([#996](https://github.com/NVIDIA-NeMo/Guardrails/issues/996))
- Add SHA-256 hashing option ([#988](https://github.com/NVIDIA-NeMo/Guardrails/issues/988))
- Add Fiddler Guardrails integration ([#964](https://github.com/NVIDIA-NeMo/Guardrails/issues/964), [#1043](https://github.com/NVIDIA-NeMo/Guardrails/issues/1043))
- Add generation metadata to streaming chunks ([#1011](https://github.com/NVIDIA-NeMo/Guardrails/issues/1011))
- Improve alpha to beta bot migration ([#878](https://github.com/NVIDIA-NeMo/Guardrails/issues/878))
- Support multimodal input and output rails ([#1033](https://github.com/NVIDIA-NeMo/Guardrails/issues/1033))
- Add support for NemoGuard JailbreakDetect NIM.  ([#1038](https://github.com/NVIDIA-NeMo/Guardrails/issues/1038))
- Set default start and end reasoning tokens ([#1050](https://github.com/NVIDIA-NeMo/Guardrails/issues/1050))
- Improve output rails error handling for SSE format ([#1058](https://github.com/NVIDIA-NeMo/Guardrails/issues/1058))

### 🐛 Bug Fixes

- Ensure parse_task_output is called after all llm_call invocations ([#1047](https://github.com/NVIDIA-NeMo/Guardrails/issues/1047))
- Handle exceptions in generate_events to propagate errors in streaming ([#1012](https://github.com/NVIDIA-NeMo/Guardrails/issues/1012))
- Ensure output rails streaming is enabled explicitly ([#1045](https://github.com/NVIDIA-NeMo/Guardrails/issues/1045))
- Improve multimodal prompt length calculation for base64 images ([#1053](https://github.com/NVIDIA-NeMo/Guardrails/issues/1053))

### 🚜 Refactor

- Move startup and shutdown logic to lifespan in server  ([#999](https://github.com/NVIDIA-NeMo/Guardrails/issues/999))

### 📚 Documentation

- Add multimodal rails documentation ([#1061](https://github.com/NVIDIA-NeMo/Guardrails/issues/1061))
- Add content safety tutorial ([#1042](https://github.com/NVIDIA-NeMo/Guardrails/issues/1042))
- Revise reasoning model info ([#1062](https://github.com/NVIDIA-NeMo/Guardrails/issues/1062))
- Consider new GS experience ([#1005](https://github.com/NVIDIA-NeMo/Guardrails/issues/1005))
- Restore deleted configuration files ([#963](https://github.com/NVIDIA-NeMo/Guardrails/issues/963))

### ⚙️ Miscellaneous Tasks

- Add Python 3.12 support ([#984](https://github.com/NVIDIA-NeMo/Guardrails/issues/984))

## [0.12.0] - 2025-02-26

### 🚀 Features

- Support Output Rails Streaming ([#966](https://github.com/NVIDIA-NeMo/Guardrails/issues/966), [#1003](https://github.com/NVIDIA-NeMo/Guardrails/issues/1003))
- Add unified output mapping for actions ([#965](https://github.com/NVIDIA-NeMo/Guardrails/issues/965))
- Add output rails support to activefence integration ([#940](https://github.com/NVIDIA-NeMo/Guardrails/issues/940))
- Add Prompt Security integration ([#920](https://github.com/NVIDIA-NeMo/Guardrails/issues/920))
- Add pii masking capability to PrivateAI integration ([#901](https://github.com/NVIDIA-NeMo/Guardrails/issues/901))
- Add embedding_params to BasicEmbeddingsIndex ([#898](https://github.com/NVIDIA-NeMo/Guardrails/issues/898))
- Add score threshold to AnalyzerEngine ([#845](https://github.com/NVIDIA-NeMo/Guardrails/issues/845))

### 🐛 Bug Fixes

- Fix dependency resolution issues in AlignScore Dockerfile([#1002](https://github.com/NVIDIA-NeMo/Guardrails/issues/1002), [#982](https://github.com/NVIDIA-NeMo/Guardrails/issues/982))
- Fix JailbreakDetect docker files([#981](https://github.com/NVIDIA-NeMo/Guardrails/issues/981), [#1001](https://github.com/NVIDIA-NeMo/Guardrails/pull/1001))
- Fix TypeError from attempting to unpack already-unpacked dictionary. ([#959](https://github.com/NVIDIA-NeMo/Guardrails/issues/959))
- Fix token stats usage in LLM call info. ([#953](https://github.com/NVIDIA-NeMo/Guardrails/issues/953))
- Handle unescaped quotes in generate_value using safe_eval ([#946](https://github.com/NVIDIA-NeMo/Guardrails/issues/946))
- Handle non-relative file paths ([#897](https://github.com/NVIDIA-NeMo/Guardrails/issues/897))
- Set workdir to models and specify entrypoint explicitly ([#1001](https://github.com/NVIDIA-NeMo/Guardrails/pull/1001)).

### 📚 Documentation

- Output streaming ([#976](https://github.com/NVIDIA-NeMo/Guardrails/issues/976))
- Fix typos with oauthtoken ([#957](https://github.com/NVIDIA-NeMo/Guardrails/issues/957))
- Fix broken link in prompt security ([#978](https://github.com/NVIDIA-NeMo/Guardrails/issues/978))
- Update advanced user guides per v0.11.1 doc release ([#937](https://github.com/NVIDIA-NeMo/Guardrails/issues/937))

### ⚙️ Miscellaneous Tasks

- Tolerate prompt in code blocks ([#1004](https://github.com/NVIDIA-NeMo/Guardrails/issues/1004))
- Update YAML indent to use two spaces ([#1009](https://github.com/NVIDIA-NeMo/Guardrails/issues/1009))

## [0.11.1] - 2025-01-16

### Added

- **ContentSafety**: Add ContentSafety NIM connector ([#930](https://github.com/NVIDIA-NeMo/Guardrails/pull/930)) by @prasoonvarshney
- **TopicControl**: Add TopicControl NIM connector ([#930](https://github.com/NVIDIA-NeMo/Guardrails/pull/930)) by @makeshn
- **JailbreakDetect**: Add jailbreak detection NIM connector ([#930](https://github.com/NVIDIA-NeMo/Guardrails/pull/930)) by @erickgalinkin

## Changed

- **AutoAlign Integration**: Add further enhancements and refactoring to AutoAlign integration ([#867](https://github.com/NVIDIA-NeMo/Guardrails/pull/867)) by @KimiJL

## Fixed

- **PrivateAI Integration**: Fix Incomplete URL substring sanitization Error ([#883](https://github.com/NVIDIA-NeMo/Guardrails/pull/883)) by @NJ-186

## Documentation

- **NVIDIA Blueprint**: Add Safeguarding AI Virtual Assistant NIM Blueprint NemoGuard NIMs ([#932](https://github.com/NVIDIA-NeMo/Guardrails/pull/932)) by @abodhankar

- **ActiveFence Integration**: Fix flow definition in community docs ([#890](https://github.com/NVIDIA-NeMo/Guardrails/pull/890)) by @noamlevy81

## [0.11.0] - 2024-11-19

### Added

- **Observability**: Add observability support with support for different backends ([#844](https://github.com/NVIDIA-NeMo/Guardrails/pull/844)) by @Pouyanpi
- **Private AI Integration**: Add Private AI Integration ([#815](https://github.com/NVIDIA-NeMo/Guardrails/pull/815)) by @letmerecall
- **Patronus Evaluate API Integration**: Patronus Evaluate API Integration ([#834](https://github.com/NVIDIA-NeMo/Guardrails/pull/834)) by @varjoshi
- **railsignore**: Add support for .railsignore file ([#790](https://github.com/NVIDIA-NeMo/Guardrails/pull/790)) by @ajanitshimanga

### Changed

- **Sandboxed Environment in Jinja2**: Add sandboxed environment in Jinja2 ([#799](https://github.com/NVIDIA-NeMo/Guardrails/pull/799)) by @Pouyanpi
- **Langchain 3 support**: Upgrade LangChain to Version 0.3 ([#784](https://github.com/NVIDIA-NeMo/Guardrails/pull/784)) by @Pouyanpi
- **Python 3.8**: Drop support for Python 3.8 ([#803](https://github.com/NVIDIA-NeMo/Guardrails/pull/803)) by @Pouyanpi
- **vllm**: Bump vllm from 0.2.7 to 0.5.5 for llama_guard and patronusai([#836](https://github.com/NVIDIA-NeMo/Guardrails/pull/836))

### Fixed

- **Guardrails Library documentation**": Fix a typo in guardrails library documentation ([#793](https://github.com/NVIDIA-NeMo/Guardrails/pull/793)) by @vedantnaik19
- **Contributing Guide**: Fix incorrect folder name & pre-commit setup in CONTRIBUTING.md ([#800](https://github.com/NVIDIA-NeMo/Guardrails/pull/800))
- **Contributing Guide**: Added correct Python command version in documentation([#801](https://github.com/NVIDIA-NeMo/Guardrails/pull/801)) by @ravinder-tw
- **retrieve chunk action**: Fix presence of new line in retrieve chunk action ([#809](https://github.com/NVIDIA-NeMo/Guardrails/pull/809)) by @Pouyanpi
- **Standard Library import**: Fix guardrails standard library import path in Colang 2.0 ([#835](https://github.com/NVIDIA-NeMo/Guardrails/pull/835)) by @Pouyanpi
- **AlignScore Dockerfile**: Add nltk's punkt_tab in align_score Dockerfile ([#841](https://github.com/NVIDIA-NeMo/Guardrails/pull/841)) by @yonromai
- **Eval dependencies**: Make pandas version constraint explicit for eval optional dependency ([#847](https://github.com/NVIDIA-NeMo/Guardrails/pull/847)) by @Pouyanpi
- **tests**: Mock PromptSession to prevent console error ([#851](https://github.com/NVIDIA-NeMo/Guardrails/pull/851)) by @Pouyanpi
- **Streaming*: Handle multiple output parsers in generation ([#854](https://github.com/NVIDIA-NeMo/Guardrails/pull/854)) by @Pouyanpi

### Documentation

- **User Guide**: Update role from bot to assistant ([#852](https://github.com/NVIDIA-NeMo/Guardrails/pull/852)) by @Pouyanpi
- **Installation Guide**: Update optional dependencies install ([#853](https://github.com/NVIDIA-NeMo/Guardrails/pull/853)) by @Pouyanpi
- **Documentation Restructuring**: Restructure the docs and several style enhancements ([#855](https://github.com/NVIDIA-NeMo/Guardrails/pull/855)) by @Pouyanpi
- **Got It AI deprecation**: Add deprecation notice for Got It AI integration ([#857](https://github.com/NVIDIA-NeMo/Guardrails/pull/857)) by @mlmonk

## [0.10.1] - 2024-10-02

- Colang 2.0-beta.4 patch

## [0.10.0] - 2024-09-27

### Added

- **content safety**: Implement content safety module ([#674](https://github.com/NVIDIA-NeMo/Guardrails/pull/674)) by @Pouyanpi
- **migration tool**: Enhance migration tool capabilities ([#624](https://github.com/NVIDIA-NeMo/Guardrails/pull/624)) by @Pouyanpi
- **Cleanlab Integration**: Add Cleanlab's Trustworthiness Score ([#572](https://github.com/NVIDIA-NeMo/Guardrails/pull/572)) by @AshishSardana
- **Colang 2**: LLM chat interface development ([#709](https://github.com/NVIDIA-NeMo/Guardrails/pull/709)) by @schuellc-nvidia
- **embeddings**: Add relevant chunk support to Colang 2 ([#708](https://github.com/NVIDIA-NeMo/Guardrails/pull/708)) by @Pouyanpi
- **library**: Migrate Cleanlab to Colang 2 and add exception handling ([#714](https://github.com/NVIDIA-NeMo/Guardrails/pull/714)) by @Pouyanpi
- **Colang debug library**: Develop debugging tools for Colang ([#560](https://github.com/NVIDIA-NeMo/Guardrails/pull/560)) by @schuellc-nvidia
- **debug CLI**: Extend debugging command-line interface ([#717](https://github.com/NVIDIA-NeMo/Guardrails/pull/717)) by @schuellc-nvidia
- **embeddings**: Add support for embeddings only with search threshold ([#733](https://github.com/NVIDIA-NeMo/Guardrails/pull/733)) by @Pouyanpi
- **embeddings**: Add embedding-only support to Colang 2 ([#737](https://github.com/NVIDIA-NeMo/Guardrails/pull/737)) by @Pouyanpi
- **embeddings**: Add relevant chunks prompts ([#745](https://github.com/NVIDIA-NeMo/Guardrails/pull/745)) by @Pouyanpi
- **gcp moderation**: Implement GCP-based moderation tools ([#727](https://github.com/NVIDIA-NeMo/Guardrails/pull/727)) by @kauabh
- **migration tool**: Sample conversation syntax conversion ([#764](https://github.com/NVIDIA-NeMo/Guardrails/pull/764)) by @Pouyanpi
- **llmrails**: Add serialization support for LLMRails ([#627](https://github.com/NVIDIA-NeMo/Guardrails/pull/627)) by @Pouyanpi
- **exceptions**: Initial support for exception handling ([#384](https://github.com/NVIDIA-NeMo/Guardrails/pull/384)) by @drazvan
- **evaluation tooling**: Develop new evaluation tools ([#677](https://github.com/NVIDIA-NeMo/Guardrails/pull/677)) by @drazvan
- **Eval UI**: Add support for tags in the Evaluation UI ([#731](https://github.com/NVIDIA-NeMo/Guardrails/pull/731)) by @drazvan
- **guardrails library**: Launch Colang 2.0 Guardrails Library ([#689](https://github.com/NVIDIA-NeMo/Guardrails/pull/689)) by @drazvan
- **configuration**: Revert abc bot to Colang v1 and separate v2 configuration ([#698](https://github.com/NVIDIA-NeMo/Guardrails/pull/698)) by @drazvan

### Changed

- **api**: Update Pydantic validators ([#688](https://github.com/NVIDIA-NeMo/Guardrails/pull/688)) by @Pouyanpi
- **standard library**: Refactor and migrate standard library components ([#625](https://github.com/NVIDIA-NeMo/Guardrails/pull/625)) by @Pouyanpi

- Upgrade langchain-core and jinja2 dependencies ([#766](https://github.com/NVIDIA-NeMo/Guardrails/pull/766)) by @Pouyanpi

### Fixed

- **documentation**: Fix broken links ([#670](https://github.com/NVIDIA-NeMo/Guardrails/pull/670)) by @buvnswrn
- **hallucination-check**: Correct hallucination-check functionality ([#679](https://github.com/NVIDIA-NeMo/Guardrails/pull/679)) by @Pouyanpi
- **streaming**: Fix NVIDIA AI endpoints streaming issues ([#654](https://github.com/NVIDIA-NeMo/Guardrails/pull/654)) by @Pouyanpi
- **hallucination-check**: Resolve non-OpenAI hallucination check issue ([#681](https://github.com/NVIDIA-NeMo/Guardrails/pull/681)) by @Pouyanpi
- **import error**: Fix Streamlit import error ([#686](https://github.com/NVIDIA-NeMo/Guardrails/pull/686)) by @Pouyanpi
- **prompt override**: Fix override prompt self-check facts ([#621](https://github.com/NVIDIA-NeMo/Guardrails/pull/621)) by @Pouyanpi
- **output parser**: Resolve deprecation warning in output parser ([#691](https://github.com/NVIDIA-NeMo/Guardrails/pull/691)) by @Pouyanpi
- **patch**: Fix langchain_nvidia_ai_endpoints patch ([#697](https://github.com/NVIDIA-NeMo/Guardrails/pull/697)) by @Pouyanpi
- **runtime issues**: Address Colang 2 runtime issues ([#699](https://github.com/NVIDIA-NeMo/Guardrails/pull/699)) by @schuellc-nvidia
- **send event**: Change 'send event' to 'send' ([#701](https://github.com/NVIDIA-NeMo/Guardrails/pull/701)) by @Pouyanpi
- **output parser**: Fix output parser validation ([#704](https://github.com/NVIDIA-NeMo/Guardrails/pull/704)) by @Pouyanpi
- **passthrough_fn**: Pass config and kwargs to passthrough_fn runnable ([#695](https://github.com/NVIDIA-NeMo/Guardrails/pull/695)) by @vpr1995
- **rails exception**: Fix rails exception migration ([#705](https://github.com/NVIDIA-NeMo/Guardrails/pull/705)) by @Pouyanpi
- **migration**: Replace hyphens and apostrophes in migration ([#725](https://github.com/NVIDIA-NeMo/Guardrails/pull/725)) by @Pouyanpi
- **flow generation**: Fix LLM flow continuation generation ([#724](https://github.com/NVIDIA-NeMo/Guardrails/pull/724)) by @schuellc-nvidia
- **server command**: Fix CLI server command ([#723](https://github.com/NVIDIA-NeMo/Guardrails/pull/723)) by @Pouyanpi
- **embeddings filesystem**: Fix cache embeddings filesystem ([#722](https://github.com/NVIDIA-NeMo/Guardrails/pull/722)) by @Pouyanpi
- **outgoing events**: Process all outgoing events ([#732](https://github.com/NVIDIA-NeMo/Guardrails/pull/732)) by @sklinglernv
- **generate_flow**: Fix a small bug in the generate_flow action for Colang 2 ([#710](https://github.com/NVIDIA-NeMo/Guardrails/pull/710)) by @drazvan
- **triggering flow id**: Fix the detection of the triggering flow id ([#728](https://github.com/NVIDIA-NeMo/Guardrails/pull/728)) by @drazvan
- **LLM output**: Fix multiline LLM output syntax error for dynamic flow generation ([#748](https://github.com/NVIDIA-NeMo/Guardrails/pull/748)) by @radinshayanfar
- **scene form**: Fix the scene form and choice flows in the Colang 2 standard library ([#741](https://github.com/NVIDIA-NeMo/Guardrails/pull/741)) by @sklinglernv

### Documentation

- **Cleanlab**: Update community documentation for Cleanlab integration ([#713](https://github.com/NVIDIA-NeMo/Guardrails/pull/713)) by @Pouyanpi
- **rails exception handling**: Add notes for Rails exception handling in Colang 2.x ([#744](https://github.com/NVIDIA-NeMo/Guardrails/pull/744)) by @Pouyanpi
- **LLM per task**: Document LLM per task functionality ([#676](https://github.com/NVIDIA-NeMo/Guardrails/pull/676)) by @Pouyanpi

### Others

- **relevant_chunks**: Add the `relevant_chunks` to the GPT-3.5 general prompt template ([#678](https://github.com/NVIDIA-NeMo/Guardrails/pull/678)) by @drazvan
- **flow names**: Ensure flow names don't start with keywords ([#637](https://github.com/NVIDIA-NeMo/Guardrails/pull/637)) by @schuellc-nvidia

## [0.9.1.1] - 2024-07-26

### Fixed

- [#650](https://github.com/NVIDIA-NeMo/Guardrails/pull/650) Fix gpt-3.5-turbo-instruct prompts #651.

## [0.9.1] - 2024-07-25

### Added

- Colang version [2.0-beta.2](./CHANGELOG-Colang.md#20-beta2---unreleased)
- [#370](https://github.com/NVIDIA-NeMo/Guardrails/pull/370) Add Got It AI's Truthchecking service for RAG applications by @mlmonk.
- [#543](https://github.com/NVIDIA-NeMo/Guardrails/pull/543) Integrating AutoAlign's guardrail library with NeMo Guardrails by @abhijitpal1247.
- [#566](https://github.com/NVIDIA-NeMo/Guardrails/pull/566) Autoalign factcheck examples by @abhijitpal1247.
- [#518](https://github.com/NVIDIA-NeMo/Guardrails/pull/518) Docs: add example config for using models with ollama by @vedantnaik19.
- [#538](https://github.com/NVIDIA-NeMo/Guardrails/pull/538) Support for `--default-config-id` in the server.
- [#539](https://github.com/NVIDIA-NeMo/Guardrails/pull/539) Support for `LLMCallException`.
- [#548](https://github.com/NVIDIA-NeMo/Guardrails/pull/548) Support for custom embedding models.
- [#617](https://github.com/NVIDIA-NeMo/Guardrails/pull/617) NVIDIA AI Endpoints embeddings.
- [#462](https://github.com/NVIDIA-NeMo/Guardrails/pull/462) Support for calling embedding models from langchain-nvidia-ai-endpoints.
- [#622](https://github.com/NVIDIA-NeMo/Guardrails/pull/622) Patronus Lynx Integration.

### Changed

- [#597](https://github.com/NVIDIA-NeMo/Guardrails/pull/597) Make UUID generation predictable in debug-mode.
- [#603](https://github.com/NVIDIA-NeMo/Guardrails/pull/603) Improve chat cli logging.
- [#551](https://github.com/NVIDIA-NeMo/Guardrails/pull/551) Upgrade to Langchain 0.2.x by @nicoloboschi.
- [#611](https://github.com/NVIDIA-NeMo/Guardrails/pull/611) Change default templates.
- [#545](https://github.com/NVIDIA-NeMo/Guardrails/pull/545) NVIDIA API Catalog and NIM documentation update.
- [#463](https://github.com/NVIDIA-NeMo/Guardrails/pull/463) Do not store pip cache during docker build by @don-attilio.
- [#629](https://github.com/NVIDIA-NeMo/Guardrails/pull/629) Move community docs to separate folder.
- [#647](https://github.com/NVIDIA-NeMo/Guardrails/pull/647) Documentation updates.
- [#648](https://github.com/NVIDIA-NeMo/Guardrails/pull/648) Prompt improvements for Llama-3 models.

### Fixed

- [#482](https://github.com/NVIDIA-NeMo/Guardrails/pull/482) Update README.md by @curefatih.
- [#530](https://github.com/NVIDIA-NeMo/Guardrails/pull/530) Improve the test serialization test to make it more robust.
- [#570](https://github.com/NVIDIA-NeMo/Guardrails/pull/570) Add support for FacialGestureBotAction by @elisam0.
- [#550](https://github.com/NVIDIA-NeMo/Guardrails/pull/550) Fix issue #335 - make import errors visible.
- [#547](https://github.com/NVIDIA-NeMo/Guardrails/pull/547) Fix LLMParams bug and add unit tests (fixes #158).
- [#537](https://github.com/NVIDIA-NeMo/Guardrails/pull/537) Fix directory traversal bug.
- [#536](https://github.com/NVIDIA-NeMo/Guardrails/pull/536) Fix issue #304 NeMo Guardrails packaging.
- [#539](https://github.com/NVIDIA-NeMo/Guardrails/pull/539) Fix bug related to the flow abort logic in Colang 1.0 runtime.
- [#612](https://github.com/NVIDIA-NeMo/Guardrails/pull/612) Follow-up fixes for the default prompt change.
- [#585](https://github.com/NVIDIA-NeMo/Guardrails/pull/585) Fix Colang 2.0 state serialization issue.
- [#486](https://github.com/NVIDIA-NeMo/Guardrails/pull/486) Fix select model type and custom prompts task.py by @cyun9601.
- [#487](https://github.com/NVIDIA-NeMo/Guardrails/pull/487) Fix custom prompts configuration manual.md.
- [#479](https://github.com/NVIDIA-NeMo/Guardrails/pull/479) Fix static method and classmethod action decorators by @piotrm0.
- [#544](https://github.com/NVIDIA-NeMo/Guardrails/pull/544) Fix issue #216 bot utterance.
- [#616](https://github.com/NVIDIA-NeMo/Guardrails/pull/616) Various fixes.
- [#623](https://github.com/NVIDIA-NeMo/Guardrails/pull/623) Fix path traversal check.

## [0.9.0] - 2024-05-08

### Added

- [Colang 2.0 Documentation](https://docs.nvidia.com/nemo/guardrails/colang-2/overview.html).
- Revamped [NeMo Guardrails Documentation](https://docs.nvidia.com/nemo-guardrails).

### Fixed

- [#461](https://github.com/NVIDIA-NeMo/Guardrails/pull/461) Feature/ccl cleanup.
- [#483](https://github.com/NVIDIA-NeMo/Guardrails/pull/483) Fix dictionary expression evaluation bug.
- [#467](https://github.com/NVIDIA-NeMo/Guardrails/pull/467) Feature/colang doc related cleanups.
- [#484](https://github.com/NVIDIA-NeMo/Guardrails/pull/484) Enable parsing of `..."<NLD>"` expressions.
- [#478](https://github.com/NVIDIA-NeMo/Guardrails/pull/478) Fix #420 - evaluate not working with chat models.

## [0.8.3] - 2024-04-18

### Changed

- [#453](https://github.com/NVIDIA-NeMo/Guardrails/pull/453) Update documentation for NVIDIA API Catalog example.

### Fixed

- [#382](https://github.com/NVIDIA-NeMo/Guardrails/pull/382) Fix issue with `lowest_temperature` in self-check and hallucination rails.
- [#454](https://github.com/NVIDIA-NeMo/Guardrails/pull/454) Redo fix for #385.
- [#442](https://github.com/NVIDIA-NeMo/Guardrails/pull/442) Fix README type by @dileepbapat.

## [0.8.2] - 2024-04-01

### Added

- [#402](https://github.com/NVIDIA-NeMo/Guardrails/pull/402) Integrate Vertex AI Models into Guardrails by @aishwaryap.
- [#403](https://github.com/NVIDIA-NeMo/Guardrails/pull/403) Add support for NVIDIA AI Endpoints by @patriciapampanelli
- [#396](https://github.com/NVIDIA-NeMo/Guardrails/pull/396) Docs/examples nv ai foundation models.
- [#438](https://github.com/NVIDIA-NeMo/Guardrails/pull/438) Add research roadmap documentation.

### Changed

- [#389](https://github.com/NVIDIA-NeMo/Guardrails/pull/389) Expose the `verbose` parameter through `RunnableRails` by @d-mariano.
- [#415](https://github.com/NVIDIA-NeMo/Guardrails/pull/415) Enable `print(...)` and `log(...)`.
- [#389](https://github.com/NVIDIA-NeMo/Guardrails/pull/389) Expose verbose arg in RunnableRails by @d-mariano.
- [#414](https://github.com/NVIDIA-NeMo/Guardrails/pull/414) Feature/colang march release.
- [#416](https://github.com/NVIDIA-NeMo/Guardrails/pull/416) Refactor and improve the verbose/debug mode.
- [#418](https://github.com/NVIDIA-NeMo/Guardrails/pull/418) Feature/colang flow context sharing.
- [#425](https://github.com/NVIDIA-NeMo/Guardrails/pull/425) Feature/colang meta decorator.
- [#427](https://github.com/NVIDIA-NeMo/Guardrails/pull/427) Feature/colang single flow activation.
- [#426](https://github.com/NVIDIA-NeMo/Guardrails/pull/426) Feature/colang 2.0 tutorial.
- [#428](https://github.com/NVIDIA-NeMo/Guardrails/pull/428) Feature/Standard library and examples.
- [#431](https://github.com/NVIDIA-NeMo/Guardrails/pull/431) Feature/colang various improvements.
- [#433](https://github.com/NVIDIA-NeMo/Guardrails/pull/433) Feature/Colang 2.0 improvements: generate_async support, stateful API.

### Fixed

- [#412](https://github.com/NVIDIA-NeMo/Guardrails/pull/412) Fix #411 - explain rails not working for chat models.
- [#413](https://github.com/NVIDIA-NeMo/Guardrails/pull/413) Typo fix: Comment in llm_flows.co by @habanoz.
- [#420](https://github.com/NVIDIA-NeMo/Guardrails/pull/430) Fix typo for hallucination message.

## [0.8.1] - 2024-03-15

### Added

- [#377](https://github.com/NVIDIA-NeMo/Guardrails/pull/377) Add example for streaming from custom action.

### Changed

- [#380](https://github.com/NVIDIA-NeMo/Guardrails/pull/380) Update installation guide for OpenAI usage.
- [#401](https://github.com/NVIDIA-NeMo/Guardrails/pull/401) Replace YAML import with new import statement in multi-modal example.

### Fixed

- [#398](https://github.com/NVIDIA-NeMo/Guardrails/pull/398) Colang parser fixes and improvements.
- [#394](https://github.com/NVIDIA-NeMo/Guardrails/pull/394) Fixes and improvements for Colang 2.0 runtime.
- [#381](https://github.com/NVIDIA-NeMo/Guardrails/pull/381) Fix typo by @serhatgktp.
- [#379](https://github.com/NVIDIA-NeMo/Guardrails/pull/379) Fix missing prompt in verbose mode for chat models.
- [#400](https://github.com/NVIDIA-NeMo/Guardrails/pull/400) Fix Authorization header showing up in logs for NeMo LLM.

## [0.8.0] - 2024-02-28

### Added

- [#292](https://github.com/NVIDIA-NeMo/Guardrails/pull/292) [Jailbreak heuristics](./docs/getting-started/tutorials/jailbreak-detection-heuristics.mdx) by @erickgalinkin.
- [#256](https://github.com/NVIDIA-NeMo/Guardrails/pull/256) Support [generation options](./docs/run-rails/using-python-apis/generation-options.mdx).
- [#307](https://github.com/NVIDIA-NeMo/Guardrails/pull/307) Added support for multi-config api calls by @makeshn.
- [#293](https://github.com/NVIDIA-NeMo/Guardrails/pull/293) Adds configurable stop tokens by @zmackie.
- [#334](https://github.com/NVIDIA-NeMo/Guardrails/pull/334) Colang 2.0 - Preview by @schuellc.
- [#208](https://github.com/NVIDIA-NeMo/Guardrails/pull/208) Implement cache embeddings (resolves #200) by @Pouyanpi.
- [#331](https://github.com/NVIDIA-NeMo/Guardrails/pull/331) Huggingface pipeline streaming by @trebedea.

Documentation:

- [#311](https://github.com/NVIDIA-NeMo/Guardrails/pull/311) Update documentation to demonstrate the use of output rails when using a custom RAG by @niels-garve.
- [#347](https://github.com/NVIDIA-NeMo/Guardrails/pull/347) Add [detailed logging docs](./docs/observability/logging/index.mdx) by @erickgalinkin.
- [#354](https://github.com/NVIDIA-NeMo/Guardrails/pull/354) [Input and output rails only guide](./docs/run-rails/using-python-apis/check-messages.mdx) by @trebedea.
- [#359](https://github.com/NVIDIA-NeMo/Guardrails/pull/359) Added [user guide for jailbreak detection heuristics](./docs/getting-started/tutorials/jailbreak-detection-heuristics.mdx) by @makeshn.
- [#363](https://github.com/NVIDIA-NeMo/Guardrails/pull/363) Add [multi-config API call user guide](./docs/run-rails/using-fastapi-server/list-guardrail-configs.mdx).
- [#297](https://github.com/NVIDIA-NeMo/Guardrails/pull/297) Example configurations for using only the guardrails, without LLM generation.

### Changed

- [#309](https://github.com/NVIDIA-NeMo/Guardrails/pull/309) Change the paper citation from ArXiV to EMNLP 2023 by @manuelciosici
- [#319](https://github.com/NVIDIA-NeMo/Guardrails/pull/319) Enable embeddings model caching.
- [#267](https://github.com/NVIDIA-NeMo/Guardrails/pull/267) Make embeddings computing async and add support for batching.
- [#281](https://github.com/NVIDIA-NeMo/Guardrails/pull/281) Follow symlinks when building knowledge base by @piotrm0.
- [#280](https://github.com/NVIDIA-NeMo/Guardrails/pull/280) Add more information to results of `retrieve_relevant_chunks` by @piotrm0.
- [#332](https://github.com/NVIDIA-NeMo/Guardrails/pull/332) Update docs for batch embedding computations.
- [#244](https://github.com/NVIDIA-NeMo/Guardrails/pull/244) Docs/edit getting started by @DougAtNvidia.
- [#333](https://github.com/NVIDIA-NeMo/Guardrails/pull/333) Follow-up to PR 244.
- [#341](https://github.com/NVIDIA-NeMo/Guardrails/pull/341) Updated 'fastembed' version to 0.2.2 by @NirantK.

### Fixed

- [#286](https://github.com/NVIDIA-NeMo/Guardrails/pull/286) Fixed #285 - using the same evaluation set given a random seed for topical rails by @trebedea.
- [#336](https://github.com/NVIDIA-NeMo/Guardrails/pull/336) Fix #320. Reuse the asyncio loop between sync calls.
- [#337](https://github.com/NVIDIA-NeMo/Guardrails/pull/337) Fix stats gathering in a parallel async setup.
- [#342](https://github.com/NVIDIA-NeMo/Guardrails/pull/342) Fixes OpenAI embeddings support.
- [#346](https://github.com/NVIDIA-NeMo/Guardrails/pull/346) Fix issues with KB embeddings cache, bot intent detection and config ids validator logic.
- [#349](https://github.com/NVIDIA-NeMo/Guardrails/pull/349) Fix multi-config bug, asyncio loop issue and cache folder for embeddings.
- [#350](https://github.com/NVIDIA-NeMo/Guardrails/pull/350) Fix the incorrect logging of an extra dialog rail.
- [#358](https://github.com/NVIDIA-NeMo/Guardrails/pull/358) Fix Openai embeddings async support.
- [#362](https://github.com/NVIDIA-NeMo/Guardrails/pull/362) Fix the issue with the server being pointed to a folder with a single config.
- [#352](https://github.com/NVIDIA-NeMo/Guardrails/pull/352) Fix a few issues related to jailbreak detection heuristics.
- [#356](https://github.com/NVIDIA-NeMo/Guardrails/pull/356) Redo followlinks PR in new code by @piotrm0.

## [0.7.1] - 2024-02-01

### Changed

- [#288](https://github.com/NVIDIA-NeMo/Guardrails/pull/288) Replace SentenceTransformers with FastEmbed.

## [0.7.0] - 2024-01-31

### Added

- [#254](https://github.com/NVIDIA-NeMo/Guardrails/pull/254) Support for [Llama Guard input and output content moderation](./docs/configure-rails/guardrail-catalog/content-safety.mdx#llama-guard-based-content-moderation).
- [#253](https://github.com/NVIDIA-NeMo/Guardrails/pull/253) Support for [server-side threads](./docs/run-rails/using-fastapi-server/overview.mdx).
- [#235](https://github.com/NVIDIA-NeMo/Guardrails/pull/235) Improved [LangChain integration](./docs/integration/langchain/langchain-integration.mdx) through `RunnableRails`.
- [#190](https://github.com/NVIDIA-NeMo/Guardrails/pull/190) Add [example](./examples/notebooks/generate_events_and_streaming.ipynb) for using `generate_events_async` with streaming.
- Support for Python 3.11.

### Changed

- [#240](https://github.com/NVIDIA-NeMo/Guardrails/pull/240) Switch to pyproject.
- [#276](https://github.com/NVIDIA-NeMo/Guardrails/pull/276) Upgraded Typer to 0.9.

### Fixed

- [#286](https://github.com/NVIDIA-NeMo/Guardrails/pull/286) Fixed not having the same evaluation set given a random seed for topical rails.
- [#239](https://github.com/NVIDIA-NeMo/Guardrails/pull/239) Fixed logging issue where `verbose=true` flag did not trigger expected log output.
- [#228](https://github.com/NVIDIA-NeMo/Guardrails/pull/228) Fix docstrings for various functions.
- [#242](https://github.com/NVIDIA-NeMo/Guardrails/pull/242) Fix Azure LLM support.
- [#225](https://github.com/NVIDIA-NeMo/Guardrails/pull/225) Fix annoy import, to allow using without.
- [#209](https://github.com/NVIDIA-NeMo/Guardrails/pull/209) Fix user messages missing from prompt.
- [#261](https://github.com/NVIDIA-NeMo/Guardrails/pull/261) Fix small bug in `print_llm_calls_summary`.
- [#252](https://github.com/NVIDIA-NeMo/Guardrails/pull/252) Fixed duplicate loading for the default config.
- Fixed the dependencies pinning, allowing a wider range of dependencies versions.
- Fixed sever security issues related to uncontrolled data used in path expression and information exposure through an exception.

## [0.6.1] - 2023-12-20

### Added

- Support for `--version` flag in the CLI.

### Changed

- Upgraded `langchain` to `0.0.352`.
- Upgraded `httpx` to `0.24.1`.
- Replaced deprecated `text-davinci-003` model with `gpt-3.5-turbo-instruct`.

### Fixed

- [#191](https://github.com/NVIDIA-NeMo/Guardrails/pull/191): Fix chat generation chunk issue.

## [0.6.0] - 2023-12-13

### Added

- Support for [explicit definition](./docs/configure-rails/yaml-schema/guardrails-configuration.mdx) of input/output/retrieval rails.
- Support for [custom tasks and their prompts](./docs/configure-rails/yaml-schema/prompt-configuration.mdx).
- Support for fact-checking [using AlignScore](./docs/configure-rails/guardrail-catalog/community/alignscore.mdx).
- Support for [NeMo LLM Service](./docs/about/supported-llms.mdx) as an LLM provider.
- Support for making a single LLM call for both the guardrails process and generating the response (by setting `rails.dialog.single_call.enabled` to `True`).
- Support for [sensitive data detection](./docs/configure-rails/guardrail-catalog/community/presidio.mdx) guardrails using Presidio.
- [Example](./examples/configs/llm/hf_pipeline_llama2) using NeMo Guardrails with the LLaMa2-13B model.
- [Dockerfile](./Dockerfile) for building a Docker image.
- Support for [prompting modes](./docs/configure-rails/yaml-schema/prompt-configuration.mdx) using `prompting_mode`.
- Support for [TRT-LLM](./docs/about/supported-llms.mdx) as an LLM provider.
- Support for [streaming](./docs/run-rails/using-python-apis/streaming.mdx) the LLM responses when no output rails are used.
- [Integration](./docs/configure-rails/guardrail-catalog/community/active-fence.mdx) of ActiveFence ActiveScore API as an input rail.
- Support for `--prefix` and `--auto-reload` in the [guardrails server](./docs/run-rails/using-fastapi-server/overview.mdx).
- Example authentication dialog flow.
- Example [RAG using Pinecone](./examples/configs/rag/pinecone).
- Support for loading a configuration from dictionary, i.e. `RailsConfig.from_content(config=...)`.
- Guidance on [LLM support](./docs/about/supported-llms.mdx).
- Support for `LLMRails.explain()` (see the [Getting Started](./docs/getting-started/installation-guide.mdx) guide for sample usage).

### Changed

- Allow context data directly in the `/v1/chat/completion` using messages with the type `"role"`.
- Allow calling a subflow whose name is in a variable, e.g. `do $some_name`.
- Allow using actions which are not `async` functions.
- Disabled pretty exceptions in CLI.
- Upgraded dependencies.
- Updated the [Getting Started Guide](./docs/getting-started/installation-guide.mdx).
- Main [README](./README.md) now provides more details.
- Merged original examples into a single [ABC Bot](./examples/bots/abc) and removed the original ones.
- Documentation improvements.

### Fixed

- Fix going over the maximum prompt length using the `max_length` attribute in [Prompt Templates](./docs/configure-rails/yaml-schema/prompt-configuration.mdx).
- Fixed problem with `nest_asyncio` initialization.
- [#144](https://github.com/NVIDIA-NeMo/Guardrails/pull/144) Fixed TypeError in logging call.
- [#121](https://github.com/NVIDIA-NeMo/Guardrails/pull/109) Detect chat model using openai engine.
- [#109](https://github.com/NVIDIA-NeMo/Guardrails/pull/109) Fixed minor logging issue.
- Parallel flow support.
- Fix `HuggingFacePipeline` bug related to LangChain version upgrade.

## [0.5.0] - 2023-09-04

### Added

- Support for [custom configuration data](./docs/configure-rails/custom-initialization/custom-data.mdx).
- Example for using custom LLM and multiple KBs.
- Support for [`PROMPTS_DIR`](./docs/configure-rails/yaml-schema/prompt-configuration.mdx).
- [#101](https://github.com/NVIDIA-NeMo/Guardrails/pull/101) Support for [using OpenAI embeddings](./docs/configure-rails/custom-initialization/custom-embedding-providers.mdx) models in addition to SentenceTransformers.
- First set of end-to-end QA tests for the example configurations.
- Support for configurable [embedding search providers](./docs/configure-rails/other-configurations/embedding-search-providers.mdx)

### Changed

- Moved to using `nest_asyncio` for [implementing the blocking API](./docs/run-rails/using-python-apis/core-classes.mdx). Fixes [#3](https://github.com/NVIDIA-NeMo/Guardrails/issues/3) and [#32](https://github.com/NVIDIA-NeMo/Guardrails/issues/32).
- Improved event property validation in `new_event_dict`.
- Refactored imports to allow installing from source without Annoy/SentenceTransformers (would need a custom embedding search provider to work).

### Fixed

- Fixed when the `init` function from `config.py` is called to allow custom LLM providers to be registered inside.
- [#93](https://github.com/NVIDIA-NeMo/Guardrails/pull/93): Removed redundant `hasattr` check in `nemoguardrails/llm/params.py`.
- [#91](https://github.com/NVIDIA-NeMo/Guardrails/issues/91): Fixed how default context variables are initialized.

## [0.4.0] - 2023-08-03

### Added

- [Event-based API](./docs/run-rails/using-python-apis/event-based-api.mdx) for guardrails.
- Support for message with type "event" in [`LLMRails.generate_async`](/guardrails-python-sdk/nemoguardrails/rails/llm/llmrails#nemoguardrails-rails-llm-llmrails-LLMRails).
- Support for [bot message instructions](./docs/configure-rails/colang/usage-examples/bot-message-instructions.mdx).
- Support for [using variables inside bot message definitions](./docs/configure-rails/colang/colang-1/colang-language-syntax-guide.mdx#bot-messages-with-variables).
- Support for `vicuna-7b-v1.3` and `mpt-7b-instruct`.
- Topical evaluation results for `vicuna-7b-v1.3` and `mpt-7b-instruct`.
- Support to use different models for different LLM tasks.
- Support for [red-teaming](./docs/evaluation/llm-vulnerability-scanning.mdx) using challenges.
- Support to disable the Chat UI when running the server using `--disable-chat-ui`.
- Support for accessing the API request headers in server mode.
- Support to [enable CORS settings](./docs/run-rails/using-fastapi-server/overview.mdx) for the guardrails server.

### Changed

- Changed the naming of the internal events to align to the upcoming UMIM spec (Unified Multimodal Interaction Management).
- If there are no user message examples, the bot messages examples lookup is disabled as well.

### Fixed

- [#58](https://github.com/NVIDIA-NeMo/Guardrails/issues/58): Fix install on Mac OS 13.
- [#55](https://github.com/NVIDIA-NeMo/Guardrails/issues/55): Fix bug in example causing config.py to crash on computers with no CUDA-enabled GPUs.
- Fixed the model name initialization for LLMs that use the `model` kwarg.
- Fixed the Cohere prompt templates.
- [#55](https://github.com/NVIDIA-NeMo/Guardrails/issues/83): Fix bug related to LangChain callbacks initialization.
- Fixed generation of "..." on value generation.
- Fixed the parameters type conversion when invoking actions from Colang (previously everything was string).
- Fixed `model_kwargs` property for the `WrapperLLM`.
- Fixed bug when `stop` was used inside flows.
- Fixed Chat UI bug when an invalid guardrails configuration was used.

## [0.3.0] - 2023-06-30

### Added

- Support for defining [subflows](./docs/configure-rails/colang/colang-1/colang-language-syntax-guide.mdx#subflows).
- Improved support for [customizing LLM prompts](./docs/configure-rails/yaml-schema/prompt-configuration.mdx)
  - Support for using filters to change how variables are included in a prompt template.
  - Output parsers for prompt templates.
  - The `verbose_v1` formatter and output parser to be used for smaller models that don't understand Colang very well in a few-shot manner.
  - Support for including context variables in prompt templates.
  - Support for chat models i.e. prompting with a sequence of messages.
- Experimental support for allowing the LLM to generate [multi-step flows](./docs/configure-rails/yaml-schema/guardrails-configuration.mdx).
- Example of using Llama Index from a guardrails configuration (#40).
- [Example](examples/configs/llm/hf_endpoint) for using HuggingFace Endpoint LLMs with a guardrails configuration.
- [Example](examples/configs/llm/hf_pipeline_dolly) for using HuggingFace Pipeline LLMs with a guardrails configuration.
- Support to alter LLM parameters passed as `model_kwargs` in LangChain.
- CLI tool for running evaluations on the different steps (e.g., canonical form generation, next steps, bot message) and on existing rails implementation (e.g., moderation, jailbreak, fact-checking, and hallucination).
- [Initial evaluation](./docs/evaluation/evaluate-guardrails.mdx) results for `text-davinci-003` and `gpt-3.5-turbo`.
- The `lowest_temperature` can be set through the guardrails config (to be used for deterministic tasks).

### Changed

- The core templates now use Jinja2 as the rendering engines.
- Improved the internal prompting architecture, now using an LLM Task Manager.

### Fixed

- Fixed bug related to invoking a chain with multiple output keys.
- Fixed bug related to tracking the output stats.
- #51: Bug fix - avoid str concat with None when logging user_intent.
- #54: Fix UTF-8 encoding issue and add embedding model configuration.

## [0.2.0] - 2023-05-31

### Added

- Support to [connect any LLM](./docs/about/supported-llms.mdx) that implements the BaseLanguageModel interface from  LangChain.
- Support for [customizing the prompts](./docs/configure-rails/yaml-schema/prompt-configuration.mdx) for specific LLM models.
- Support for [custom initialization](./docs/configure-rails/custom-initialization/index.mdx) when loading a configuration through `config.py`.
- Support to extract [user-provided values](./docs/configure-rails/colang/usage-examples/extract-user-provided-values.mdx) from utterances.

### Changed

- Improved the logging output for Chat CLI (clear events stream, prompts, completion, timing information).
- Updated system actions to use temperature 0 where it makes sense, e.g., canonical form generation, next step generation, fact checking, etc.
- Excluded the default system flows from the "next step generation" prompt.
- Updated langchain to 0.0.167.

### Fixed

- Fixed initialization of LangChain tools.
- Fixed the overriding of general instructions [#7](https://github.com/NVIDIA-NeMo/Guardrails/issues/7).
- Fixed action parameters inspection bug [#2](https://github.com/NVIDIA-NeMo/Guardrails/issues/2).
- Fixed bug related to multi-turn flows [#13](https://github.com/NVIDIA-NeMo/Guardrails/issues/13).
- Fixed Wolfram Alpha error reporting in the sample execution rail.

## [0.1.0] - 2023-04-25

### Added

- First alpha release.