How jibankumar-cloud/faber works
Faber is a terminal-based AI coding assistant. The principal workflow is: a developer configures a model, submits a task, the agent streams model responses and dispatches tools, and approved edits or shell commands change the workspace; checkpoints support undo. Its code graph, persistent memory, session history, and usage ledger support that loop. The graph follows the supplied excerpts, SOURCE INDEX references, and README; unsampled implementation details and their unverified connections are not inferred.
Main entry points: agent.ts Streaming agent loop · llm.ts Model API client · index.ts Terminal REPL · config.ts Runtime configuration
Components 24
CLI and setup
- Terminal REPL
src/index.ts - Guided setup
src/onboard.ts - Runtime configuration
src/config.ts - Profile settings
src/settings.ts
Agent workflow
- Streaming agent loop
src/agent.ts
Project state
- Code graph index
src/indexer.ts - Long-term memory
src/memory/longTerm.ts - Conversation context
src/memory/shortTerm.ts - Session history
src/memory/sessions.ts - Usage ledger
src/usage.ts
Model providers
- Provider routes
src/routes.ts - Credential resolution
src/credentials.ts - Model API client
src/llm.ts - AWS request signing
src/sigv4.ts - Model catalog
src/models.ts - Model pricing
src/pricing.ts
Safe execution
- Tool dispatch
src/tools/registry.ts - Workspace file tools
src/tools/fs.ts - Shell execution
src/tools/shell.ts - Change checkpoints
src/checkpoints.ts - Task git commits
src/git.ts
Other components
- Developer
- Project workspace
- Model provider
Connections 31
- Developer → Terminal REPL: submits task
README.md - Terminal REPL → Guided setup: starts setup
src/index.ts - Terminal REPL → Runtime configuration: loads config
src/index.ts - Runtime configuration → Profile settings: loads profiles
src/config.ts - Runtime configuration → Provider routes: selects route
src/config.ts - Runtime configuration → Credential resolution: resolves keys
src/config.ts - Guided setup → Provider routes: lists providers
src/onboard.ts - Terminal REPL → Streaming agent loop: runs tasks
src/index.ts - Streaming agent loop → Model API client: requests responses
src/agent.ts - Streaming agent loop → Tool dispatch: dispatches tools
src/agent.ts - Streaming agent loop → Change checkpoints: creates checkpoints
src/agent.ts - Streaming agent loop → Long-term memory: uses memory
src/agent.ts - Streaming agent loop → Conversation context: uses context
src/agent.ts - Streaming agent loop → Session history: persists sessions
src/agent.ts - Streaming agent loop → Code graph index: queries code graph
src/agent.ts - Streaming agent loop → Usage ledger: records usage
src/agent.ts - Streaming agent loop → Task git commits: commits task
src/agent.ts - Tool dispatch → Workspace file tools: dispatches file tools
src/tools/registry.ts - Tool dispatch → Shell execution: dispatches commands
src/tools/registry.ts - Workspace file tools → Project workspace: reads and edits
src/tools/fs.ts - Shell execution → Project workspace: runs commands
README.md - Workspace file tools → Change checkpoints: records changes
src/tools/fs.ts - Model API client → Model provider: streams requests
src/llm.ts - Model API client → AWS request signing: signs AWS requests
src/llm.ts - Model API client → Provider routes: uses route config
src/llm.ts - Model API client → Model catalog: uses model metadata
src/llm.ts - Model API client → Model pricing: uses pricing data
src/llm.ts - Usage ledger → Model pricing: costs tokens
src/usage.ts - Developer → Workspace file tools: approves edits
README.md - Developer → Shell execution: approves commands
README.md - Developer → Terminal REPL: undoes tasks
README.md