API Guide

This guide documents JD.Writer API behavior for local development and integrations.

Base URL (default local): http://localhost:19081 (E2E/test host) or http://localhost:5319 (direct dotnet run)

Endpoints

POST /ai/continue

Generates continuation text for a draft.

Request body:

{
  "draft": "# Draft title\nExisting content..."
}

Response body:

{
  "continuation": "\n\n## Next Steps\n...",
  "source": "ollama"
}

POST /ai/slash

Executes a slash command transform.

Request body:

{
  "command": "summarize",
  "draft": "Long markdown content...",
  "prompt": "Optional custom prompt override"
}

Response body:

{
  "output": "Short summary...",
  "source": "native-llama-cpu"
}

POST /ai/assist/stream

Streams NDJSON insight chunks for side panels.

Request body:

{
  "mode": "hints",
  "draft": "Current note text...",
  "prompt": "Optional panel prompt override"
}

Response content type:

application/x-ndjson

Example chunks:

{"text":"Consider adding an explicit goal statement.","source":"ollama"}
{"text":"Split this section into short bullets for readability.","source":"native-llama-gpu"}

GET /ai/provider-summary

Returns active provider preference and readiness diagnostics:

  • openai configured status
  • ollama model/endpoint readiness inputs
  • native llama chain configuration
  • hardware scan summary (GPU names + preferred acceleration)
  • active runtime provider order

Provider Configuration

Primary configuration file: JD.Writer.ApiService/appsettings.json

Key settings:

  • AI:Provider: auto, ollama, native, local, openai, fallback
  • AI:Ollama:Endpoint: Ollama URL
  • AI:Ollama:Model: model name
  • AI:OpenAI:ApiKey: OpenAI API key (optional)
  • AI:NativeLlama:Enabled: enable native llama chain (default true)
  • AI:NativeLlama:CliPath: full path to llama-cli/main executable
  • AI:NativeLlama:ModelDirectory: optional directory scanned for .gguf models
  • AI:NativeLlama:GpuModelPath: GGUF model path for GPU-capable runs
  • AI:NativeLlama:CpuModelPath: GGUF model path for CPU fallback runs
  • AI:NativeLlama:ContextSize, MaxTokens, Threads, GpuLayers, Temperature, TimeoutSeconds

Auto-discovery behavior:

  • If CliPath is empty, runtime probes PATH for llama-cli, main, or llamafile.
  • If GpuModelPath/CpuModelPath are empty, runtime probes ModelDirectory then common local roots for .gguf.

Curl Quick Checks

curl -sS http://localhost:5319/ai/provider-summary

curl -sS -X POST http://localhost:5319/ai/continue \
  -H "Content-Type: application/json" \
  -d "{\"draft\":\"Draft intro\"}"

curl -N -X POST http://localhost:5319/ai/assist/stream \
  -H "Content-Type: application/json" \
  -d "{\"mode\":\"hints\",\"draft\":\"Draft intro\"}"

Reference API Docs

For generated code-level API documentation, see API Reference.

Edit this page