Version: v0.0

Origin finds possible direct source candidates for a thought.

It owns one job:

check whether a specific retained source closely matches or introduced a thought

Origin is stricter than Influence. It should return fewer results and avoid weak claims.

• Accepts a user thought or question.
• Reads meaningful Inference/Memory evidence.
• Scores direct source candidates.
• Prioritizes exact phrase overlap, rare term overlap, and retained evidence strength.
• Emits guarded origin candidates with source evidence.

{
  "schema_version": "memact.origin.v0",
  "thought": "I need to build something real before applying anywhere",
  "candidates": [
    {
      "id": "act_1",
      "packet_id": "packet:act_1",
      "score": 0.91,
      "claim_type": "origin_candidate",
      "reason": "close phrase overlap with the thought and matching source evidence"
    }
  ],
  "language_guardrail": "Origin means possible source candidate, not proof."
}

Prerequisites:

• Node.js 20+
• npm 10+

Install:

npm install

Validate:

npm run check

Run sample:

npm run sample

Run against Inference output:

npm run origin -- --input path\to\inference-output.json --thought "I need to build something real" --format report

JSON output:

npm run origin -- --input path\to\inference-output.json --thought "I need to build something real" --format json

• Origin is a candidate engine, not a proof engine.
• It should prefer precision over recall.
• It should cite retained evidence.
• It should not read Capture internals.

MIT. See LICENSE.