https://github.com/sagebynature/signal-graph
CLI-first trading research toolkit for turning market events into explainable trade hypotheses with explicit provenance.
https://github.com/sagebynature/signal-graph
cli knowledge-graph market-research neo4j provenance python sqlite trading
Last synced: 19 days ago
JSON representation
CLI-first trading research toolkit for turning market events into explainable trade hypotheses with explicit provenance.
- Host: GitHub
- URL: https://github.com/sagebynature/signal-graph
- Owner: sagebynature
- Created: 2026-03-20T04:35:39.000Z (3 months ago)
- Default Branch: main
- Last Pushed: 2026-04-12T18:25:11.000Z (2 months ago)
- Last Synced: 2026-04-12T20:14:52.728Z (2 months ago)
- Topics: cli, knowledge-graph, market-research, neo4j, provenance, python, sqlite, trading
- Language: Python
- Size: 463 KB
- Stars: 1
- Watchers: 0
- Forks: 0
- Open Issues: 0
-
Metadata Files:
- Readme: README.md
Awesome Lists containing this project
README
# Signal Graph
`Signal Graph` is a local, CLI-first **memory and decision-support system** for humans and AI tools.
It captures provenance-rich signals, journals them into a graph-backed path, recalls prior activity with deterministic artifacts, and exposes the same runtime through CLI and MCP surfaces.
## Supported Command Surface
Primary user-facing commands:
- `signal-graph doctor`
- `signal-graph init`
- `signal-graph bootstrap-describe`
- `signal-graph automation-describe`
- `signal-graph capture-signal`
- `signal-graph journalize-signal`
- `signal-graph recall-signal`
- `signal-graph integration-install`
- `signal-graph integration-audit`
- `signal-graph integration-uninstall`
- `signal-graph mcp-server`
- `signal-graph-mcp`
## What Signal Graph Does
- capture owner- and actor-scoped signals with explicit provenance
- preserve raw evidence and artifact references locally
- journal signals into a graph-backed path for later explanation
- recall prior work by query, session, source, or runtime
- expose the same system through a local MCP transport for tool integrations
## Quick Start
### 1. Verify prerequisites
```bash
uv run signal-graph doctor --json
```
### 2. Initialize local state
```bash
uv run signal-graph init
```
This creates `.signal-graph/cache/`, `.signal-graph/artifacts/`, and `.signal-graph/signal_graph.db`.
### 3. Inspect the bootstrap contract
```bash
uv run signal-graph bootstrap-describe --format markdown
```
The bootstrap contract is the canonical runtime guide for supported entrypoints, smoke steps, and MCP expectations.
### 4. Capture, journal, and recall a signal
```bash
uv run signal-graph capture-signal --text "Agent completed the release checklist signal." --origin-type agent_artifact --source-name codex --process codex --runtime-family codex --session-id session-a --intent-status explicit --why "Safely finish the release." --what release --where notes/release.md
uv run signal-graph journalize-signal --signal SIGNAL_ID
uv run signal-graph recall-signal --query release
```
### 5. Inspect automation and host integration guidance
```bash
uv run signal-graph automation-describe --format markdown
uv run signal-graph integration-install --host claude-code
uv run signal-graph integration-audit --host claude-code --json
```
## MCP Runtime
`Signal Graph` ships a published stdio MCP entrypoint and an equivalent CLI launch path:
- `signal-graph-mcp`
- `signal-graph mcp-server`
Host integration examples and validation evidence live under `docs/integrations/`.
## Documentation Map
- `docs/README.md` — documentation guide and reading paths
- `docs/runbooks/operator-guide.md` — runtime, bootstrap, automation, and troubleshooting guidance
- `docs/overview/product.md` — product framing and supported use cases
- `docs/architecture/system-overview.md` — architecture boundaries and storage/transport model
- `docs/integrations/README.md` — MCP host matrix and installation examples
## Current Scope
Signal Graph is intentionally local-first. It favors explicit commands, deterministic artifacts, and inspectable provenance over hidden automation or a dashboard-first UI.
Non-goals for this cut:
- opaque autonomous decision-making
- public multi-tenant hosting
- mutating raw truth when later corrections arrive
- unsupported legacy workflow compatibility