Debugging with Weaver
Weaver’s debugging tools keep those hiccups from becoming hair-pullers.
TL;DR
1. Recognise the Red Flags
During a run, components signal status:
| Border Colour | Meaning | What to Do |
|---|---|---|
| 🟢 Green | Success | Celebrate |
| 🟡 Yellow | Waiting | Upstream input missing |
| 🔴 Red | Failed | Time to debug |
2. Inspect Inputs & Outputs
Click the 🔴 component. Weaver shows:
- Inputs it received
- Outputs / error it produced
Look for obvious culprits:
- Empty strings?
- 500-level API errors?
- Rate-limit messages?
3. Fix with AI (The 1-Click Wonder)
If the error is text-related (bad prompt, wrong params):
Article text exceeded 5 000 chars — trim to 800 and retry.
Press Fix with AI. Weaver rewrites the skill’s internal prompt or parameters, re-runs the single block, and shows the diff.
If it turns green, you’re done.
If still red, refine the instruction and try again.
4. Multi-Select for Chain Re-runs
Hold Cmd/Ctrl and click multiple components to:
- Re-execute them in order
- Debug cascading failures (e.g., bad JSON bubbles downstream)
5. Common Error Recipes
| Symptom | Likely Fix |
|---|---|
| “JSON parse error” | Add Constraints: return pure JSON in upstream skill. |
| OpenAI quota hit | Insert “Delay 60 s” skill or upgrade plan. |
6. Prevent Tomorrow’s Bugs
- Small skills > monoliths — easier to pin-point.
- Use test prompts — keep a gallery of real-world inputs.
- Watch run logs — agent activity logs surface slow steps early.
Red borders happen.
Weaver’s debug tools turn mystery errors into 3-click fixes.
7. When All Else Fails
- Press Control + Shift + E to export your AI agent instantly. Run it locally using our cross-platform runtime (Mac, Windows, Linux)
- Visit #debugging-help on Discord.
- Save your sanity.
INFO
Enable Debug Mode
Flip the Debug toggle in the top toolbar to log:
- Component timelines
- Step through each component or run an entire workflow
- Request/response payloads
- Runtime metadata (headers, params)
- Structured inputs and outputs
INFO
Reading Logs & Output
When Debug Mode is enabled, Weaver captures detailed logs for every agent run — across all components.
What You’ll See
- Timeline — Execution order of each block
- Payload — Raw input/output JSON per step
- Headers — Full HTTP headers, user agent, IPs, custom flags
- Parsed Inputs — Cleaned query/body/params data
- Component Logs — Structured logs like prompts, scrapers, APIs
- Model Info — LLM name, prompt, and reply
- Jump to Block — Clickable trace into the visual workflow
Example Log Snapshot
=== Web Scrape Log ===
Payload: {
"searchTopic": "general",
"includeQAs": true
}
=== LLM Prompt Log ===
Model: gpt-4o
Prompt: Analyze the following search results...
Reply: { "primary_keywords": [...], ... }
Component Status Palette
| Colour / Banner | Meaning | Action |
|---|---|---|
| 🟢 Green | Success | Celebrate |
| 🟡 Yellow | Waiting for upstream | Check prior block |
| 🟠 Orange banner | Missing required input | Drag a wire |
| ⚪ Grey (running) | In progress | Sit tight |
| 🔴 Red | Failed | Click → Fix with AI |
INFO