# Paste under llm-pi-ai.providers in ~/.dsh/settings.yaml (WSL).
# 1) Run host_reach (profile=all) from a DSH session and copy a reachable baseURL.
# 2) Set a dummy key in the shell that starts dsh web, e.g.:
#      export OLLAMA_API_KEY=ollama
#      export LMSTUDIO_API_KEY=lmstudio
#      export VLLM_API_KEY=vllm
#      export LLAMA_API_KEY=llama
# 3) Model ids must match what the server lists (ollama list / LM Studio / vLLM / llama-server).
# 4) On Windows, bind the server beyond 127.0.0.1 if WSL cannot reach it
#    (OLLAMA_HOST=0.0.0.0:11434, LM Studio "Serve on local network", etc.).

llm-pi-ai:
  providers:
    ollama:
      displayName: Local Ollama (Windows)
      apiKeyEnv: OLLAMA_API_KEY
      api: openai-completions
      # Replace with host_reach suggestedBaseURL for ollama (port 11434)
      baseURL: http://127.0.0.1:11434/v1
      compat:
        supportsDeveloperRole: false
        maxTokensField: max_tokens
      models:
        - id: qwen38-27b-local:latest
          name: Qwen3.8 27B local (UD-Q4)
          # Must match Ollama n_ctx (ollama show). 131072 here with n_ctx=8192 causes API errors.
          contextWindow: 8192
          maxTokens: 8192

    lmstudio:
      displayName: LM Studio (Windows)
      apiKeyEnv: LMSTUDIO_API_KEY
      api: openai-completions
      # Default LM Studio local server port
      baseURL: http://127.0.0.1:1234/v1
      compat:
        supportsDeveloperRole: false
        maxTokensField: max_tokens
      models:
        - id: REPLACE_WITH_LMSTUDIO_MODEL_ID
          name: LM Studio model
          contextWindow: 32768
          maxTokens: 8192

    vllm:
      displayName: vLLM (Docker / WSL)
      apiKeyEnv: VLLM_API_KEY
      api: openai-completions
      # Official OpenAI-compatible server often listens on 8000
      baseURL: http://127.0.0.1:8000/v1
      compat:
        supportsDeveloperRole: false
        maxTokensField: max_tokens
      models:
        - id: REPLACE_WITH_VLLM_SERVED_MODEL
          name: vLLM served model
          contextWindow: 131072
          maxTokens: 8192

    llama:
      displayName: llama-server / Unsloth Desktop
      apiKeyEnv: LLAMA_API_KEY
      api: openai-completions
      # Confirm port in Unsloth Desktop / llama-server UI (8080 is common)
      baseURL: http://127.0.0.1:8080/v1
      compat:
        supportsDeveloperRole: false
        maxTokensField: max_tokens
      models:
        - id: REPLACE_WITH_LLAMA_SERVER_MODEL
          name: llama-server model
          contextWindow: 131072
          maxTokens: 8192