classify

Classify Input

Send arbitrary text or content to The Hoardinator's classification router (classify-input Edge Function). Returns structured intelligence: flags, summary, action items, memories, and best quotes.

Prerequisites

• SUPABASE_PROJECT_REF environment variable set (hdhmwaldvzxwhimoemap)
• SUPABASE_ACCESS_TOKEN or SUPABASE_ANON_KEY environment variable set
• Content to classify (pasted text, a file, or the current conversation context)

Step 1: Determine What to Classify

Accept content from:

• Explicit argument: /classify <text> — use the provided text directly
• Clipboard/file: If user says "classify this file" or pastes content, use that
• Current context: If no argument, summarize what's being discussed and classify it

Step 2: Create an Input Record

First insert the content into public.inputs so the classifier can process it:

curl -s -X POST \
  "https://api.supabase.com/v1/projects/${SUPABASE_PROJECT_REF}/database/query" \
  -H "Authorization: Bearer ${SUPABASE_ACCESS_TOKEN}" \
  -H "Content-Type: application/json" \
  -d '{
    "query": "INSERT INTO public.inputs (type, status, data) VALUES ('\''note'\'', '\''ingested'\'', '\''{\"source_type\": \"manual\", \"content\": \"<ESCAPED_CONTENT>\"}'\''::jsonb) RETURNING id"
  }'

Capture the returned id as INPUT_ID.

Step 3: Call classify-input Edge Function

curl -s -X POST \
  "https://${SUPABASE_PROJECT_REF}.supabase.co/functions/v1/classify-input" \
  -H "Authorization: Bearer ${SUPABASE_ANON_KEY}" \
  -H "Content-Type: application/json" \
  -d "{\"input_id\": \"${INPUT_ID}\"}"

The response includes:

{
  "classified_id": "ops_clsf_XXXXX",
  "flags": ["jtbd", "story", "action-item"],
  "summary": "Two-sentence summary of what was classified."
}

Step 4: Fetch Full Classification

To get the full structured output (action items, memories, quotes):

curl -s -X POST \
  "https://api.supabase.com/v1/projects/${SUPABASE_PROJECT_REF}/database/query" \
  -H "Authorization: Bearer ${SUPABASE_ACCESS_TOKEN}" \
  -H "Content-Type: application/json" \
  -d "{\"query\": \"SELECT data FROM classified WHERE id = '${CLASSIFIED_ID}'\"}"

Step 5: Report Results

Present to the user:

## Classification Results

**Summary:** <2-sentence summary>

**Flags:** jtbd, story, action-item (+ N others)

**Action Items:**
- [ ] <task> (@owner)

**Memories Captured:**
- <person>: <durable fact>

**Best Quotes:**
> "<quote>" — <speaker>

**Classified ID:** ops_clsf_XXXXX (stored in Hoardinator)

Important Notes

• Content is stored permanently in Supabase after classification
• action-item flags automatically create dispatch_queue entries for downstream agents
• The classifier uses 25 flags: jtbd, story, quote, deal, outreach, proposal, asset-needed, idea, framework, expert, product-gap, pricing, market, channel, memory, permission, action-item, coaching, delivery, partnership, and others