# LexPipe ai-assistant

Streaming chat orchestration. Routes ``search_matters`` tool calls to postgres-dsl-assistant, persists threads / files via the .NET app, and runs LLM completions through the configured providers below.

## API docs

• /docs (Swagger UI)
• /redoc (ReDoc)
• /openapi.json

## Probes

• /live
• /ready

## LLM providers

One row per provider this service can talk to.

Provider	Env var	Status
OpenAI	OPENAI_API_KEY	✓ set
X.ai	XAI_API_KEY	✓ set
NVIDIA	NVIDIA_API_KEY	✓ set
Anthropic	ANTHROPIC_API_KEY	✓ set
Xiaomi Mimo	MIMO_API_KEY	✗ unset
Google Gemini	GOOGLE_GEMINI_API_KEY	✓ set

## Chat settings

Setting	Value
Max steps (tool loop)	10
Max output tokens / turn	4096
Max parallel tool calls	5
Max tool calls per step	10
Model override	(unset — per-request default)

## Backends

Backend	Value
.NET host (auth + account/info)	https://main-app-prod--0000021.wittyglacier-a6d709c1.eastus.azurecontainerapps.io
Chat CRUD host	https://main-app-prod--0000021.wittyglacier-a6d709c1.eastus.azurecontainerapps.io (fallback)
Matter search host	https://dsl-assistant-prod--4dr4p5g.wittyglacier-a6d709c1.eastus.azurecontainerapps.io
Assistant secret	set