Terminal interface for the Bonfires AI API. Chat with agents, search the knowledge graph, sync context, and render graph data — all from the command line.
pip install bonfiresOr install from source:
pip install git+https://github.com/NERDDAO/bonfire-cli.gitFor global install without polluting your Python environment:
pipx install bonfiresRun the interactive setup:
bonfire initThis will:
- Connect to the Bonfires API
- List your bonfires and let you pick one
- List available agents and let you pick one
- Save credentials to
~/.config/bonfires/config.env
You're ready to go:
bonfire chat "hello"Interactive setup wizard. Connects to the API, lists your bonfires and agents, and saves config.
bonfire initOr skip the prompts:
bonfire init --api-key YOUR_KEY --bonfire-id ID --agent-id IDSend a message to a Bonfire agent. Searches the knowledge graph by default.
bonfire chat "What do we know about auth patterns?"Control graph behavior with --graph-mode:
bonfire chat "find related context" # default: regenerate
bonfire chat --graph-mode adaptive "let the agent decide"
bonfire chat --graph-mode append "add to existing graph"
bonfire chat --graph-mode static "just reply, no graph"Search the knowledge graph directly.
bonfire delve "error handling patterns"
bonfire delve -n 20 "authentication architecture"Push context to the knowledge graph. Derives a chat ID from your current git repo and branch.
bonfire sync "shipped bonfire-cli v0.1"
bonfire sync "design spec for auth" -f docs/auth-design.mdList agents for the configured bonfire.
List all bonfires.
Render graph data from a JSON file or stdin.
bonfire graph results.json
cat graph.json | bonfire graphFormat raw API responses piped from stdin — useful for scripting.
curl -s ... | bonfire format-chatManage kEngrams — verifiable knowledge subgraphs. Each kEngram is a curated projection of the canonical Bonfires KG with cryptographic provenance: every pinned entity gets a SHA-256 fingerprint of its content at pin time, rolled into a merkle root.
bonfire kengram new "Session Name" # create and set active
bonfire kengram pin <uuid> # pin entity (auto-fetches from KG)
bonfire kengram pin --search "rubric pipeline" # search KG, pick from results
bonfire kengram create "Entity" --summary "…" # push new entity to KG + pin
bonfire kengram verify # check against canonical KG
bonfire kengram verify --local # local-only integrity check
bonfire kengram delete <id> --force # remove a kEngram
bonfire kengram show # show active kEngram
bonfire kengram merge <src> --into <tgt> # merge session → topic
bonfire kengram export # export to Obsidian .canvas
bonfire kengram list # list allHow verification works: verify batch-fetches all pinned entities from the canonical KG, recomputes their hashes from current KG content, and compares against the stored hashes. Per-entity status: OK (unchanged), DRIFT (KG content changed since pin), NOT IN KG (entity deleted). If the API is unreachable, falls back to local-only integrity checking.
Pin modes: Provide a UUID to auto-fetch entity metadata from the KG. Use --search to find entities interactively. Use --name/--summary/--labels for manual metadata (fallback if API is down).
Set BONFIRE_VAULT_DIR to control where manifests and canvas files are stored (default: ~/Vaults/Bonfires/vault).
Config is loaded in this order (later overrides earlier):
~/.config/bonfires/config.env(created bybonfire init).envin the current directory- Environment variables
| Variable | Description |
|---|---|
BONFIRE_API_URL |
API base URL (default: https://tnt-v2.api.bonfires.ai) |
BONFIRE_ID |
Your bonfire ID |
BONFIRE_AGENT_ID |
Agent ID to interact with |
BONFIRE_API_KEY |
API key for authentication |
BONFIRE_VAULT_DIR |
kEngram storage directory (default: ~/Vaults/Bonfires/vault) |