Gong Processor

Chrome returns raw transcript text + metadata only. All analysis and writes happen in Claude.

Local transcript files

Naming: CLIENTYYMMDD.txt

• Date: YYMMDD → 20YY-MM-DD
• Client: everything before the 6-digit date
• gong_call_id = filename without extension (e.g., bakker260303)

Step 2 — Load Aliases + Clients

curl -s "{supabase_url}/rest/v1/aliases?select=*" \
  -H "apikey: {KEY}" -H "Authorization: Bearer {KEY}"

curl -s "{supabase_url}/rest/v1/clients?select=id,name" \
  -H "apikey: {KEY}" -H "Authorization: Bearer {KEY}"

Apply alias corrections to every client name, advisor name, and project name encountered.

Step 3 — Client Identification

1. Use Gong "Company" field or filename prefix
2. Apply alias corrections from Step 2
3. Fuzzy-match corrected name against all client slugs/names
4. ≥85% → proceed with that client_id
5. 60–84% → flag as unmatched_client; do not write; skip call
6. <60% → flag as unmatched_client; do not write; skip call

Step 4 — Full Analysis (gong-review framework)

Run all 7 sections for every call. Do not skip or abbreviate.

4.1 Sentiment (per speaker)

• Tone: positive / neutral / negative / mixed
• Energy: engaged / passive / distracted / resistant
• Key emotional signals with brief characterization (transcript-grounded)
• Weight client speaker statements more heavily than advisor statements
• Default to "neutral" — never default to "positive" without evidence

4.2 Projects Mentioned

For each distinct project: name (client's language), status (NEW / ongoing), short description (NEW only), who mentioned it.

Strong signals (create/match): explicit "project" label, named initiative, campaign name, proper noun + action noun. Not a project: one-time tasks, recurring processes, past events.

4.3 Tasks Per Project

High-confidence task patterns:

• "[Name] is going to / will [verb]..." → task for that advisor
• "We need to [verb]..." → task for call participant advisor
• "I'll [verb]..." → task for speaking advisor
• "Action item:", "Follow up on...", "Send...", "Draft...", "Review..." → tasks

Do not create tasks for: completed past actions ("we sent that last week"), hypotheticals ("we could potentially..."), client commitments that are not advisor actions.

4.4 MOI Scorecard (8 categories, baseline 4.0, cap 6.9)

Score each with: score (float), rationale (transcript-grounded), deduction_triggers (what cost points).

1. Strategic Value Delivered — Did the advisor move meaningful work forward?
2. Clarity & Communication — Were action items, timelines, and expectations communicated crisply?
3. Client Engagement & Buy-In — Did the client actively engage? Was buy-in secured?
4. Follow-Up Readiness — Were specific next steps and owners defined?
5. Depth vs. Breadth — Was time spent on high-impact topics vs. scattered across surface issues?
6. Meaning — Did the advisor connect the work to the client's larger goals?
7. Outcome — Was there a clear outcome or decision from this call?
8. Impact — Will this call measurably improve the client's trajectory?

Overall MOI = average of 8 scores, rounded to 1 decimal.

Deduction triggers (apply −0.2 to −0.6 per instance):

• Vague action items without owner or date
• Client confusion or unresolved questions
• Advisor talking past client signals
• Missed opportunity to deepen strategic discussion

Excellence signals (apply +0.2 to +0.6 per instance):

• Specific, time-bound commitments secured from client
• Advisor identified a risk or opportunity the client hadn't seen
• Clear energy shift — client went from skeptical to engaged
• Advisor proactively connected dots across multiple projects

4.5 Reflection Questions (3 required — transcript evidence mandatory)

1. What was the single most important moment in this call, and why?
2. Where did the advisor create the most value — and where did they leave value on the table?
3. If you could replay one minute of this call differently, what would it be and why?

4.6 Strategic Summary

2–3 sentences: call strengths and top improvement opportunities. Factual, no editorializing. Written from the perspective of briefing a team lead who wasn't on the call.

4.7 Top 3 Core Values Demonstrated

Choose from: Inspirational, Disciplined, Accountable, Transparent, Aligned, Results-Oriented. Evidence-based only — cite the transcript moment that demonstrates each value.

Step 5 — Deduplication (REQUIRED before any writes)

5.1 Project Dedup

# Fetch all existing projects for this client
curl -s "{supabase_url}/rest/v1/projects?client_id=eq.{slug}&select=id,name,status" \
  -H "apikey: {KEY}" -H "Authorization: Bearer {KEY}"

For each extracted project name:

• ≥85% fuzzy match → use existing project ID → route tasks to Output 2 (existing project)
• 60–84% fuzzy match → flag as ambiguous_project; still process tasks but add flag note; do not merge
• <60% match → create new project → route tasks to Output 1 (new project)

5.2 Task Dedup (per project)

# Fetch all existing tasks for this project
curl -s "{supabase_url}/rest/v1/tasks?project_id=eq.{pid}&select=id,title,completed&order=created_at.desc" \
  -H "apikey: {KEY}" -H "Authorization: Bearer {KEY}"

For each extracted task:

• >85% title similarity with PENDING task → SKIP creation; PATCH due_date only if new date is earlier
• >85% title similarity with COMPLETED task → CREATE new (recurring action)
• No match → CREATE new task

Step 6 — Build 3 Output Objects

Output 1 — New Projects + Tasks

Projects from transcript with <60% match to existing (new).

{
  "project": {
    "id": "[new uuid]",
    "client_id": "[slug]",
    "name": "[project name from transcript]",
    "status": "active",
    "adv": "[O/F/M based on advisor role]",
    "sort_order": 0
  },
  "tasks": [
    {
      "id": "[new uuid]",
      "project_id": "[project uuid above]",
      "title": "[task title]",
      "due_date": "YYYY-MM-DD or null",
      "completed": false,
      "adv": "[O/F/M]",
      "sort_order": 0,
      "gong_call_id": "[call id]",
      "gong_recording_link": "[URL or null]"
    }
  ]
}

Output 2 — New Tasks on Existing Projects

Tasks from transcript where project matched ≥85%.

{
  "project_id": "[existing project uuid]",
  "project_name": "[matched project name]",
  "match_confidence": 0.92,
  "tasks": [
    {
      "id": "[new uuid]",
      "project_id": "[existing project uuid]",
      "title": "[task title]",
      "due_date": "YYYY-MM-DD or null",
      "completed": false,
      "adv": "[O/F/M]",
      "sort_order": 0,
      "gong_call_id": "[call id]",
      "gong_recording_link": "[URL or null]"
    }
  ]
}

Output 3 — MOI Call Datapoint (advisor_call_summaries)

Sentiment derived from moi_score: ≥4.0=positive, 3.0–3.9=neutral, 2.0–2.9=cautious, <2.0=negative.

{
  "client_id": "[client-slug]",
  "advisor_id": "[uuid from setup.json advisors[]]",
  "role": "operations",
  "call_date": "YYYY-MM-DD",
  "gong_call_id": "[call id or filename stem]",
  "summary": "[2-4 sentence strategic summary]",
  "key_topics": ["project name 1", "project name 2"],
  "sentiment": "positive",
  "gong_recording_link": "[URL or null]",
  "moi_score": 4.2,
  "call_summary_detail": {
    "moi_scorecard": [
      {
        "category": "Strategic Value Delivered",
        "score": 4.5,
        "rationale": "[transcript-grounded evidence]",
        "deduction_triggers": "[what cost points]"
      },
      {
        "category": "Clarity & Communication",
        "score": 4.2,
        "rationale": "...",
        "deduction_triggers": "..."
      },
      {
        "category": "Client Engagement & Buy-In",
        "score": 4.0,
        "rationale": "...",
        "deduction_triggers": "..."
      },
      {
        "category": "Follow-Up Readiness",
        "score": 4.3,
        "rationale": "...",
        "deduction_triggers": "..."
      },
      {
        "category": "Depth vs. Breadth",
        "score": 4.1,
        "rationale": "...",
        "deduction_triggers": "..."
      },
      {
        "category": "Meaning",
        "score": 4.0,
        "rationale": "...",
        "deduction_triggers": "..."
      },
      {
        "category": "Outcome",
        "score": 4.2,
        "rationale": "...",
        "deduction_triggers": "..."
      },
      {
        "category": "Impact",
        "score": 4.0,
        "rationale": "...",
        "deduction_triggers": "..."
      }
    ],
    "reflection_questions": [
      {
        "question": "What was the single most important moment in this call, and why?",
        "answer": "[specific transcript evidence]"
      },
      {
        "question": "Where did this advisor create the most value — and where did they leave value on the table?",
        "answer": "[specific transcript evidence]"
      },
      {
        "question": "If you could replay one minute of this call differently, what would it be and why?",
        "answer": "[specific transcript evidence]"
      }
    ],
    "strategic_summary": "[2-3 sentences: strengths + improvement opportunities]",
    "core_values": ["Disciplined", "Accountable", "Transparent"]
  }
}

Step 7 — Write to Supabase

Standard headers (all curl calls):

-H "apikey: {supabase_service_key}"
-H "Authorization: Bearer {supabase_service_key}"
-H "Content-Type: application/json"

Insert new projects (Output 1):

curl -s -X POST "{supabase_url}/rest/v1/projects" \
  -H "apikey: {KEY}" -H "Authorization: Bearer {KEY}" \
  -H "Content-Type: application/json" \
  -H "Prefer: resolution=merge-duplicates" \
  -d '{...project row...}'

Insert new tasks (Output 1 + Output 2, after dedup check):

curl -s -X POST "{supabase_url}/rest/v1/tasks" \
  -H "apikey: {KEY}" -H "Authorization: Bearer {KEY}" \
  -H "Content-Type: application/json" \
  -H "Prefer: resolution=merge-duplicates" \
  -d '{...task row...}'

PATCH existing task due_date (dedup: earlier date found):

curl -s -X PATCH "{supabase_url}/rest/v1/tasks?id=eq.{task_id}" \
  -H "apikey: {KEY}" -H "Authorization: Bearer {KEY}" \
  -H "Content-Type: application/json" \
  -d '{"due_date": "YYYY-MM-DD"}'

Upsert MOI call summary (Output 3, on conflict: gong_call_id → replace):

curl -s -X POST "{supabase_url}/rest/v1/advisor_call_summaries" \
  -H "apikey: {KEY}" -H "Authorization: Bearer {KEY}" \
  -H "Content-Type: application/json" \
  -H "Prefer: resolution=merge-duplicates" \
  -d '{...call summary row...}'

Recompute next_due_date for each affected project:

# Get earliest pending task with a due date
curl -s "{supabase_url}/rest/v1/tasks?project_id=eq.{pid}&completed=eq.false&due_date=not.is.null&order=due_date.asc&limit=1&select=due_date" \
  -H "apikey: {KEY}" -H "Authorization: Bearer {KEY}"

# PATCH project
curl -s -X PATCH "{supabase_url}/rest/v1/projects?id=eq.{pid}" \
  -H "apikey: {KEY}" -H "Authorization: Bearer {KEY}" \
  -H "Content-Type: application/json" \
  -d '{"next_due_date": "[date]"}'

Step 8 — Update sync_state

curl -s -X PATCH "{supabase_url}/rest/v1/sync_state?id=eq.singleton" \
  -H "apikey: {KEY}" -H "Authorization: Bearer {KEY}" \
  -H "Content-Type: application/json" \
  -d '{"last_gong_sync":"[ISO datetime]","gong_sync_in_progress":false,"gong_sync_error":null}'

If flagged items exist: read current flagged_items array, append new entries, PATCH back.

Reference Files

• references/extraction-rules.md — additional project/task extraction heuristics
• references/ambiguity-handling.md — uncertain client/project match handling
• references/document-linking.md — finding and linking Drive documents mentioned in calls