How veeresh-bikkaneti/airlock works
Airlock is a Windows-first local AI platform with two distinct user workflows: start a machine-sized chat backend, or start a separately sized coding backend and verify its tool-calling capability. The platform also offers model selection and switching, an OpenAI-compatible tool-calling proxy, and a memory service that supports retrieval-augmented chat and coding sessions. The graph includes the sampled memory-service internals; unsampled PowerShell runtime behavior is represented only where the README documents it, so some subsystem links are omitted.
Main entry points: main.py Memory API · Start-AI.ps1 Chat Startup · proxy.py Backend Forwarding · memory_store.py Vector Memory Store
Components 19
User Interfaces
- VS Code Extension
src/extension.ts - AI CLI Helpers
scripts/profile-helpers.ps1
Local Runtime
- Chat Startup
scripts/Start-AI.ps1 - Coding Session
scripts/Start-AgentSession.ps1 - Backend Capability
scripts/Get-BackendCapability.ps1 - Ollama Backend
- vLLM Backend
scripts/Start-VLLM.ps1 - llama.cpp Backend
scripts/runtime-adapters/llamacpp.ps1 - Model Registry
config/models.json
Memory Service
- Memory API
memory-service/app/main.py - Chat Retrieval Graph
memory-service/app/graph.py - Vector Memory Store
memory-service/app/memory_store.py - Embedding Backends
memory-service/app/embeddings.py - Session Checkpointer
memory-service/app/checkpointer.py - Memory Configuration
memory-service/app/config.py - Backend Forwarding
memory-service/app/proxy.py - Audit Logging
memory-service/app/audit.py
Compatibility Proxies
- Tool-Calling Proxy
tool-proxy/app/main.py
Other components
- User
Connections 25
- User → AI CLI Helpers: starts or manages
README.md - User → VS Code Extension: controls platform
README.md - VS Code Extension → AI CLI Helpers: runs helpers
src/extension.ts - AI CLI Helpers → Chat Startup: starts chat
README.md - Chat Startup → Backend Capability: checks backend
README.md - Chat Startup → Ollama Backend: starts backend
README.md - Chat Startup → vLLM Backend: may invoke
README.md - User → Coding Session: starts coding
README.md - Coding Session → llama.cpp Backend: starts backend
README.md - VS Code Extension → Model Registry: reads models
src/extension.ts - User → Tool-Calling Proxy: sends tool requests
- Memory API → Chat Retrieval Graph: orchestrates retrieval
memory-service/app/main.py - Chat Retrieval Graph → Vector Memory Store: recalls memories
memory-service/app/graph.py - Vector Memory Store → Embedding Backends: embeds documents
memory-service/app/memory_store.py - Memory API → Vector Memory Store: stores and recalls
memory-service/app/main.py - Memory API → Session Checkpointer: uses session state
memory-service/app/main.py - Memory API → Backend Forwarding: forwards chat
memory-service/app/main.py - Memory API → Audit Logging: records events
memory-service/app/main.py - Vector Memory Store → Memory Configuration: loads settings
memory-service/app/memory_store.py - Embedding Backends → Memory Configuration: reads endpoints
memory-service/app/embeddings.py - Memory API → Memory Configuration: loads paths
memory-service/app/main.py - Backend Forwarding → Ollama Backend: forwards chat
- Backend Forwarding → vLLM Backend: forwards chat
- Backend Forwarding → llama.cpp Backend: forwards coding
- Audit Logging → Memory Configuration: uses platform paths
memory-service/app/audit.py