LAB QUOTA · OK
[ retro-themes:// ] experimental
cat: ai model: @cf/meta/llama-3.1-8b-instruct

Paste retro notes (Mural / Miro stickies, Slack threads) → get consolidated themes + 3 action items the team will actually commit to.

// system prompt
You synthesise retro notes into themes + actions. User pastes raw notes.

Output format:

  ## What we heard (themes ranked by frequency)

  ### 1. <Theme name> (N mentions)
  Sentiment: <positive / negative / mixed>
  Quotes:
   - "<verbatim quote>"
   - "<verbatim quote>"
  Possible action: <one specific action>

  ### 2. ...

  ## Recommended actions for this retro (pick 3)
  1. **<concrete action>** — Owner: <role/placeholder> · By: <when> · Success measure: <how we'll know it worked>
  2. <…>
  3. <…>

  ## Items that came up but I'm parking (and why)
  • <theme> — <reason for not making it a commitment this retro>

Rules:
- Quotes must be VERBATIM. If a quote is too long, truncate with "…" but don't paraphrase.
- A theme needs ≥ 2 mentions to be ranked. Lower-frequency items go to "Parking".
- Recommended actions: exactly 3. The team can't commit to more in one retro and have it stick.
- Each action needs an owner placeholder, a deadline, and a success measure. "Improve code review" with no measure is not an action.
- If something is positive (e.g. "pair programming worked well"), recognise it — actions can be "do more of X", not just "fix Y".
⚡ powered by Cloudflare Workers AI · quota deducted on success
// sample output
## What we heard (themes ranked by frequency)

### 1. Code review turnaround (3 mentions)
Sentiment: negative
Quotes:
 - "I felt blocked on the API design for 3 days waiting on review"
 - "Code review turnaround was slow"
 - "Code reviews stuck for days again"
Possible action: Adopt a "review within 24 business hours" SLA, surfaced as a Slack reminder.

### 2. Action items vanishing (2 mentions)
Sentiment: negative
Quotes:
 - "Action items from last retro never got done"
 - "Why do action items vanish? we should track them somewhere"
Possible action: Park action items in a single tracked location (Notion / Jira board) with owner + due-by. Review at the next retro.

### 3. Standup length (2 mentions, mixed)
Sentiment: mixed (1 wants shorter, 1 wants current)
Quotes:
 - "Standup is too long, often runs 30 min"
 - "Standup is fine for me, I like the visibility"
Possible action: Trial a 15-min hard cap on standup for one sprint, with "parking lot" Slack thread for items that need more time.

### 4. Architecture / design sessions (2 mentions, positive)
Sentiment: positive
Quotes:
 - "I really enjoyed the architecture session we did Tuesday"
 - "I felt energised after the architecture session"
Possible action: Make architecture sessions a recurring slot, every 2 sprints.

## Recommended actions for this retro (pick 3)
1. **Adopt a 24-business-hour code review SLA.** Owner: tech lead. By: start of next sprint. Success measure: < 10% of PRs aged > 24 hours by sprint end.
2. **Move action-item tracking to a dedicated retro-actions board (Notion).** Owner: PM. By: end of this week. Success measure: each item has an owner + due-date; reviewed first 5 min of next retro.
3. **Trial a 15-min hard cap on standup, with Slack parking lot.** Owner: tech lead (timekeeper). By: starting tomorrow. Success measure: > 80% of standups end within 16 minutes by end of next sprint.

## Items that came up but I'm parking (and why)
• "Pair programming with Eve was great" (1 mention) — positive but single voice; flag in next retro to see if it scales.
• "Vendor docs are confusing" / "Vendor support unhelpful" (2 mentions total, related but distinct) — escalation rather than process change. Flag separately to the PM to take to the vendor.
• "New BA picking up tickets fast - really helps" (1 mention) — positive reinforcement to surface in next 1:1; not an action item.
// powered by cloudflare workers ai · quota deducted on success ← back to catalog