How microsoft/obeaver works
oBeaver is a local-first LLM inference toolkit. Engineers configure local model storage, then use interactive CLI chat or send OpenAI-compatible HTTP requests; requests are handled by a selected chat engine and streamed or returned to the caller. Separate capabilities include embeddings for retrieval, tool calling for agent workflows, model conversion, and a browser dashboard with system-resource monitoring. The source index supports several internal call relationships; other capability links below rely on README documentation, since unsampled implementation details were not inspected.
Main entry points: server.py HTTP API · cli.py CLI Commands · chat.py Terminal Chat · engine_ort.py ORT Engine
Components 14
User Interfaces
- CLI Commands
obeaver/cli.py - Terminal Chat
obeaver/chat.py - HTTP API
obeaver/server.py - Web Dashboard
obeaver/server.py
Chat Inference
- ORT Engine
obeaver/engine_ort.py - Foundry Engine
obeaver/engine_foundrylocal.py
Model Capabilities
- Tool Calling
obeaver/tools.py - Embedding Engine
obeaver/engine_embedding.py - Model Conversion
obeaver/convert_vl.py
Configuration and Monitoring
- Resource Monitor
obeaver/monitor.py - Model Configuration
obeaver/config.py
Other components
- Developer
- Foundry Local
- ONNX Runtime GenAI
Connections 19
- Developer → CLI Commands: runs commands
README.md - Developer → Terminal Chat: chats locally
README.md - Developer → HTTP API: sends requests
README.md - Developer → Web Dashboard: opens dashboard
README.md - CLI Commands → Model Configuration: configures models
obeaver/cli.py - CLI Commands → Terminal Chat: starts chat
obeaver/cli.py - Terminal Chat → ORT Engine: uses engine
obeaver/chat.py - Terminal Chat → Foundry Engine: uses engine
obeaver/chat.py - HTTP API → ORT Engine: serves inference
obeaver/server.py - HTTP API → Foundry Engine: serves inference
obeaver/server.py - Foundry Engine → Foundry Local: requests generation
obeaver/engine_foundrylocal.py - ORT Engine → ONNX Runtime GenAI: runs generation
README.md - ORT Engine → Tool Calling: uses utilities
obeaver/engine_ort.py - Foundry Engine → Tool Calling: uses utilities
obeaver/engine_foundrylocal.py - HTTP API → Embedding Engine: serves embeddings
obeaver/server.py - CLI Commands → Model Conversion: runs conversion
obeaver/cli.py - HTTP API → Resource Monitor: reads metrics
obeaver/server.py - Web Dashboard → HTTP API: uses dashboard routes
obeaver/server.py - CLI Commands → Model Configuration: reads model paths
obeaver/cli.py