CinvanaAI
← All projects

Experiment · Models & evaluation

Transcript to GitHub Projects

Study how a transcript becomes a proposed work plan while preserving the model response for review.

Repository: CinvanaAI/transcript-to-github-projects

Recorded synthetic example. An executable experiment using supplied synthetic inputs.

See the idea in action.

The actual analysis coordinator reads the input, saves the raw response and validates its project schema. The example uses injected provider/configuration boundaries and makes no network calls.

Example input
A synthetic conversation requesting a review queue and an explicitly authored provider response.
Captured result
{
  "structured_fixture": {
    "mode": "Actual coordinator; explicitly authored fixture response",
    "source": "User: Make a review queue. Require approval before publishing any item.",
    "request_messages": [
      {
        "role": "system",
        "content": "You review software-planning conversations and produce a helpful raw analysis.\n\nDo not assume a fixed schema.\nYou may answer in plain text, markdown, bullets, or JSON if you choose.\nPreserve important goals, tasks, decisions, constraints, and open questions when relevant.\nDo not invent unsupported facts."
      },
      {
        "role": "user",
        "content": "Review the following conversation transcript.\nReturn your best raw analysis for human review.\nTranscript:\nUser: Make a review queue. Require approval before publishing any item."
      }
    ],
    "captured_raw_response": "{\"project_title\": \"Workshop Review Queue\", \"project_description\": \"Review proposed work before publication.\", \"source_summary\": \"Synthetic planning request.\", \"items\": [{\"type\": \"Task\", \"title\": \"Add a review gate\", \"body\": \"Require a person to approve each draft.\", \"priority\": \"High\", \"confidence\": \"High\", \"phase\": \"V1\", \"needs_review\": true}]}",
    "raw_preserved_exactly": true,
    "schema_validation": {
      "passed": true,
      "project_title": "Workshop Review Queue"
    },
    "compatibility_status": "supported",
    "human_approval_performed": false,
    "published": false,
    "network_calls": 0
  },
  "raw_prose_fixture": {
    "mode": "Actual coordinator; explicitly authored fixture response",
    "source": "User: Make a review queue. Require approval before publishing any item.",
    "request_messages": [
      {
        "role": "system",
        "content": "You review software-planning conversations and produce a helpful raw analysis.\n\nDo not assume a fixed schema.\nYou may answer in plain text, markdown, bullets, or JSON if you choose.\nPreserve important goals, tasks, decisions, constraints, and open questions when relevant.\nDo not invent unsupported facts."
      },
      {
        "role": "user",
        "content": "Review the following conversation transcript.\nReturn your best raw analysis for human review.\nTranscript:\nUser: Make a review queue. Require approval before publishing any item."
      }
    ],
    "captured_raw_response": "The conversation asks for a review queue with explicit approval before publication.",
    "raw_preserved_exactly": true,
    "schema_validation": {
      "passed": false,
      "error_type": "JSONDecodeError"
    },
    "compatibility_status": "supported",
    "human_approval_performed": false,
    "published": false,
    "network_calls": 0
  }
}

Try the example.

From the repository root, follow the dependency requirements in the README. This example uses supplied synthetic material.

python -m pip install -r requirements.txt
python -m examples.offline_case

The actual analysis coordinator reads the input, saves the raw response and validates its project schema. The example uses injected provider/configuration boundaries and makes no network calls.

Complete setup and instructions ↗

The interesting part.

Raw model output and observed provider success are evidence to review; neither automatically authorizes public work creation.

Preserved research case

Where it came from.

Historical conversation_to_github_projects planning application.

Keeps its reusable workbench source, refreshes public imports/examples and separates its publishing helper and nested evaluator work.

The desktop publishing handoff remains deferred; a nested successor is documented without a reliable full dated chronology.

Source ↗

Follow the family: Transcript planning and reviewed GitHub publication

  • Focused extractionGitHub Project Draft Publisher

    Inspect or reuse this boundary without navigating the whole application. The parent desktop handoff is deferred; the helper has its own complete reviewed-plan interface.

    Source ↗