Skip to content

continue.md

Latest commit

 

History

History
64 lines (45 loc) · 1.79 KB

continue.md

File metadata and controls

64 lines (45 loc) · 1.79 KB

Continue.dev Integration

Inject engram's knowledge graph directly into Continue.dev chat via the @engram context provider.

Prerequisites

  • engram v0.5+ initialized in your project (engram init /path/to/project)
  • Continue.dev extension installed in VS Code or JetBrains
  • Node.js 18+

Installation

npm install engramx-continue

Configuration

Add the provider to ~/.continue/config.json:

{
  "contextProviders": [
    {
      "name": "engramx-continue"
    }
  ]
}

Restart Continue after saving the config.

Usage

In any Continue chat input, type @engram followed by your question:

@engram how does the query pipeline work?
@engram what are the known issues with the AST miner?
@engram what architectural decisions were made for the graph schema?

The provider runs engram query "<your text>" -p <workspace> --budget 2000 and injects the result as context before your message is sent to the LLM.

How It Works

  1. CLI first — calls engram query as a subprocess (5s timeout)
  2. HTTP fallback — tries http://127.0.0.1:7337/query if CLI fails (3s timeout, for Sprint 2 remote mode)
  3. Graceful degradation — returns empty context if both fail; no errors surface

Troubleshooting

@engram returns no context

  • Confirm engram is on your PATH: which engram
  • Confirm the graph is initialized: engram stats -p /path/to/project
  • Check that .engram/graph.db exists in your workspace root

graph.db not found error from CLI

  • Run engram init /path/to/project to create the initial graph
  • Then mine your codebase: engram mine /path/to/project (or just open files — the hook auto-mines)

Stale context

  • Re-index manually: engram mine /path/to/project
  • Or enable the file watcher: engram watch /path/to/project