Open-source/Self-hosted/Any LLM

Stop grepping.
Start knowing.

repi ingests logs into PostgreSQL with pgvector, clusters them by signature, and runs an autonomous ReAct loop to surface root causes.

$repi init --with-docker

Everything you need to debug faster

Built on PostgreSQL and your existing LLM provider. No new infrastructure, no vendor lock-in.

Hybrid Search

BM25 full-text + pgvector HNSW dense retrieval fused with Reciprocal Rank Fusion for best-of-both recall.

Autonomous ReAct Loop

LLM thinks, calls tools, observes results — up to 10 iterations per investigation to surface the actual root cause.

Background File Watcher

Multi-LLM

OpenAI, Anthropic, Mistral, Gemini, Ollama — swap your provider with a single environment variable.

Progressive Time Windows

Investigations start narrow and automatically expand their search window when results are sparse.

Full Audit Trail

Every thought, tool call, and observation is persisted to PostgreSQL. Replay any investigation step by step.

Up and running in 5m

Prerequisites: Docker, Python 3.11+, uv, Node.js (for the web UI).

Install dependencies

Resolves the lockfile into a local virtualenv via uv.

uv sync

Bootstrap

Starts Postgres + Redis in Docker, prompts for an LLM provider and API key, writes .env, applies the schema. Idempotent — safe to re-run.

repi init --with-docker

Start the API

Runs on http://localhost:8000. Swagger UI at /docs.

repi serve

Start the UI

Runs on http://localhost:3000 (configurable via UI_PORT). Open this in a second terminal.

repi ui

Always-on log ingestion

The worker watches directories for new log bytes and ingests them automatically. It tracks a byte offset per file so it only processes what's new — restarts are safe.

Register a watcher

curl -X POST http://localhost:8000/watchers \
  -H "Content-Type: application/json" \
  -d '{"service_name": "auth-svc", "watch_path": "/var/log/auth"}'

Start the worker

python -m repi.worker

Configuration

Set via environment variables or a .env file at the project root.

Variable	Default	Description
`REPI_ENV`	`production`	production \| development. Production is quiet (uvicorn log_level=warning, no reload). Development is verbose.
`DATABASE_URL`	`postgresql+asyncpg://lograg_user:password_here@localhost:5432/lograg`	PostgreSQL asyncpg connection URL
`LLM_PROVIDER`	`openai`	openai \| anthropic \| mistral \| gemini \| ollama
`LLM_MODEL`	`provider default`	Override the default model for the selected provider
`OPENAI_API_KEY`	`—`	Provider API key (also ANTHROPIC_API_KEY, MISTRAL_API_KEY, GEMINI_API_KEY)
`REDIS_URL`	`redis://localhost:6379`	Redis connection URL for response caching
`ENABLE_REDIS_CACHE`	`true`	Set false to disable Redis caching
`TIME_WINDOW_INITIAL_MINUTES`	`10`	Starting time window for investigations (minutes)
`TIME_WINDOW_EXPANSIONS`	`60,360,1440`	Progressive window expansion in minutes (1h → 6h → 24h)
`UI_PORT`	`3000`	Port the web UI binds to (read by `repi ui`)
`WATCHER_CONFIG_REFRESH_SECS`	`30`	How often the worker polls DB for config changes
`OLLAMA_BASE_URL`	`http://localhost:11434`	Ollama API endpoint

Investigate your first incident in under 5 minutes.

Self-hosted. No data leaves your infrastructure.