generated: '2026-08-05'
method: searched
source: https://docs.seekr.com/flow/changelog/seekr-managed/2026
also:
- https://docs.seekr.com/flow/changelog/self-hosted/2026
- https://docs.seekr.com/flow/changelog/seekr-managed/2025
- https://docs.seekr.com/flow/changelog/seekr-managed/2024
summary: >-
  Seekr publishes two parallel dated changelogs — one for the Seekr-managed (SaaS) instance and
  one for self-hosted installations. The managed changelog is release-note style, dated to the
  day, and tagged Feature / Improvement / Fix / Deprecation. The self-hosted changelog carries an
  explicit product version (e.g. 6.2026.0, 5.2.5). Each entry also marks which surfaces it lands
  on (UI / API / SDK) and, where relevant, the minimum `seekrai` SDK version required.
scheme:
  managed: dated release notes, tagged (Feature | Improvement | Fix | Deprecation)
  self_hosted: dated release notes with a product version string
  api_versioning: URI path (v1, with v2 present for the agents resource)
current:
  self_hosted_version: 6.2026.0
  self_hosted_version_date: '2026-07-17'
  sdk_version: 0.33.0
  spec_versions:
    llm-training: 5.108.1
    agents: 5.45.1
    explainability: 5.17.1
    serving: 0.1.0
archive:
  years: [2024, 2025, 2026]
entries:
- date: '2026-07-08'
  tags: [Improvement]
  highlights:
  - Trace agent instructions as a context attribution source
  - View snippets and full page content for web search sources
  - Trace agent responses to MCP tool outputs
- date: '2026-06-26'
  tags: [Feature]
  highlights:
  - Summarize the metadata in a vector database
- date: '2026-06-23'
  tags: [Improvement, Fix, Deprecation]
  breaking: true
  highlights:
  - Guide dataset generation with data job instructions
  - Track alignment job progress in real time
  - Platform improvements and fixes
  - Context-grounded fine-tuning is deprecated
  deprecation:
    feature: context-grounded fine-tuning
    announced: '2026-06-23'
    removal: '2026-07-17'
    guidance: Choose another fine-tuning method for new projects.
- date: '2026-06-22'
  tags: [Feature]
  highlights:
  - Score reinforcement tuning outputs with an LLM judge
- date: '2026-06-19'
  tags: [Feature]
  highlights:
  - Add, filter, and edit custom chunk metadata
- date: '2026-06-12'
  tags: [Feature]
  highlights:
  - Trace agent responses back to their source documents
- date: '2026-06-09'
  tags: [Feature]
  highlights:
  - Partially update agents and tools
- date: '2026-05-28'
  tags: [Improvement]
  highlights:
  - Control agent reasoning with effort levels and temperature
- date: '2026-05-21'
  tags: [Feature]
  highlights:
  - Edit agents directly in the UI
- date: '2026-05-07'
  tags: [Feature]
  highlights:
  - MCP connector support for agents
- date: '2026-04-29'
  tags: [Feature]
  highlights:
  - Role-based access control for team collaboration
- date: '2026-04-08'
  tags: [Feature]
  highlights:
  - Track file and vector database dependencies across your environment
- date: '2026-04-02'
  tags: [Feature]
  highlights:
  - Manage the full AI-ready data pipeline through a single job ID
- date: '2026-03-12'
  tags: [Feature, Improvement]
  highlights:
  - Create and manage tools from Agent Builder
  - Trace agent responses to their sources
  - Parallel tool execution for agents
- date: '2026-02'
  tags: [Feature]
  highlights:
  - Preference tuning (DPO)
  - Reinforcement tuning reward functions
- date: '2026-01'
  tags: [Feature]
  highlights:
  - Vision-language fine-tuning
  - Tools library
  - Ingestion insights
  - Multi-node fine-tuning
self_hosted_entries:
- {date: '2026-07-17', version: 6.2026.0, tags: [Improvement]}
- {date: '2026-04-22', version: 5.2.5, tags: [Feature]}
- {date: '2026-03-12', version: 5.2.5, tags: [Improvement]}
- {date: '2026-02', version: 5.0.X, tags: [Improvement]}