Run a Post-Meeting Assessment

Before you begin

Audience: Consultant, support engineer

You have completed a customer meeting and have either a list of the questions that came up during it, or a source artifact containing the meeting record that has already been ingested and has a known artifact_id. Transcripts are the primary case, but the artifact may also be structured notes or other supported text-bearing input. If you have not yet ingested the source, complete Ingest a Conversation Source first and record the artifact_id it returns.

Confirm the service is healthy before proceeding:

curl -s http://localhost:8000/health

---

Steps

1. Gather the inputs for the assessment

Decide which submission path applies to your situation:

You have	Use
A list of specific question candidate IDs (from a previous extraction)	question_candidate_ids array
An ingested source artifact	artifact_id

If you are not sure which question candidates came out of a meeting, query the intake artifact first:

curl -s "http://localhost:8000/v1/intake/artifacts/{artifact_id}/candidates"

Record the id values of the candidates you want to assess.

2. Submit the post-meeting assessment

Option A — Submit with question candidate IDs:

curl -s -X POST http://localhost:8000/v1/readout/post-meeting \
  -H "Content-Type: application/json" \
  -d '{
    "customer_scope": "acme-corp",
    "meeting_date": "2026-06-07",
    "question_candidate_ids": [
      "cand-0a1b2c3d",
      "cand-4e5f6a7b",
      "cand-8c9d0e1f"
    ]
  }'

Option B — Submit with a source artifact ID:

curl -s -X POST http://localhost:8000/v1/readout/post-meeting \
  -H "Content-Type: application/json" \
  -d '{
    "customer_scope": "acme-corp",
    "meeting_date": "2026-06-07",
    "artifact_id": "art-7f8a9b0c"
  }'

The service responds with HTTP 202 Accepted and a readout object. Record the top-level id field — this is your readout_id.

{
  "id": "rdout-3a4b5c6d",
  "status": "processing",
  "meeting_date": "2026-06-07",
  "customer_scope": "acme-corp"
}

3. Retrieve the completed assessment

Poll until status is complete:

curl -s http://localhost:8000/v1/readout/rdout-3a4b5c6d

When complete, the response includes a coverage object that categorizes every question into one of three buckets:

Category	Meaning
answered	A registry answer with sufficient relevance score was found
partially_answered	A match was found but with low confidence — review before relying on it
unanswered	No match found in the registry

4. Review partially answered matches

For every question in partially_answered, open the matched answer and check its relevance_score and answer_text directly before treating it as authoritative:

curl -s http://localhost:8000/v1/registry/answers/{answer_id}

A low relevance score means the match is approximate. The answer may address a related but different question.

Note: "Partially answered" matches may have low relevance scores. Review the source answer text and its cited evidence before relying on it in any customer-facing communication.

5. Triage unanswered questions

For each question in the unanswered list, decide on one of three follow-up actions:

Action	When to use
Submit a generation work item	Internal sources (Confluence, Jira) likely contain the answer
Escalate manually	The question requires expert input that tooling cannot produce
Defer to next cycle	Low priority; note it in your backlog

To submit a generation work item for an unanswered question, see Submit a Generation Work Item.

Important: Unanswered questions are the primary input to the AGWMP generation pipeline (Flow B). Submitting them as work items keeps the knowledge gap visible, tracked, and actionable. Questions left un-submitted remain invisible to the governance pipeline.

6. Export the assessment as Markdown

Download a formatted coverage report suitable for sharing with your team or attaching to a project record:

curl -s "http://localhost:8000/v1/readout/rdout-3a4b5c6d/export?format=markdown" \
  -o post-meeting-2026-06-07.md

The Markdown export includes the coverage breakdown, matched answer references, and the full list of unanswered questions.

Note: A PDF export option is also available. Substitute format=pdf and change the output filename accordingly. PDF output is suitable for email distribution or attaching to CRM records.

---

Result

You have a coverage report for the meeting that categorizes every question as answered, partially answered, or unanswered. Unanswered questions are documented and ready for generation pipeline submission. Partially answered matches have been reviewed for adequacy. Your team has a shared, exportable artifact showing the knowledge gaps from this meeting.