Troubleshooting
Fix generic Chat answers, Context and Live Data issues, Recipe inconsistency, Agent failures, meeting processing, Plan Mode errors, and credit limits.
Use this page when something "doesn't work like ChatGPT" — usually grounding, permissions, or plan limits.
Chat Answers Feel Generic
Symptoms: Answers could apply to any company; no citations; wrong policies.
Fixes:
- Attach a Context — picker in Chat or
@context-namein the message. - Ask for citations — "Answer only from attached sources; cite document names."
- Narrow the Context — split oversized mixed folders by job (Support vs Sales vs Legal).
- Remove noise — delete outdated drafts and duplicate uploads from the Context.
- Pick a stronger model for complex reasoning (if your plan allows).
Context / RAG Issues
| Symptom | Likely cause | Fix |
|---|---|---|
| "I don't have that information" | Content not in Context | Upload doc; check Live Source last run |
| Wrong document cited | Similar titles in same Context | Rename files; split Contexts |
| Stale answer | Static upload only | Add Live Source or re-upload |
| Processing failed on upload | OCR/size limits | Retry processing; split large PDFs |
Reprocess failed documents from the Context Documents tab when the UI offers it.
Live Data Problems
| Symptom | Fix |
|---|---|
| Live Data tab shows upgrade | Pro/Max required — Billing |
| Source paused / error | Fix URL; Run now; Resume |
| Empty after add | Wait for first ingestion; check robots.txt |
| Digest mode unexpected | Switch to Save each item for precise Q&A |
| YouTube not updating | YouTube syncs once on add — re-add if video changed |
See Live Sources.
Recipe Inconsistency
Symptoms: Same Recipe, wildly different output; wrong format; ignores Context.
Fixes:
- Add explicit output format (table, email, JSON, bullet count).
- Tighten role in the prompt ("You are a …").
- Reduce ambiguous variables — use select instead of free text where possible.
- Test with edge-case inputs (empty field, long paste, typos).
- Attach Context in Recipe settings, not only in Chat at run time.
See Recipes.
Agent and Integration Failures
| Symptom | Fix |
|---|---|
| Agent mode unavailable | Paid plan + credits |
| "Connect Gmail" (etc.) | Tools Marketplace → connect |
| Tool error mid-run | Narrow prompt; retry; check app permissions |
| SSO_REQUIRED | Enterprise SSO login |
| Recipe won't start | Connect all required toolkits |
See Agents and Integrations.
Meeting Processing
| Symptom | Fix |
|---|---|
| Bot didn't join | Valid URL; meeting allows bots; retry |
| Stuck processing | Wait; refresh page |
| No summary yet | Wait ~90s after transcript; refresh |
| Chat with Meeting empty | Wait for indexing; retry chat |
| Monthly limit | Free tier 5/month — upgrade |
Use Retry on failed meetings. See Meetings.
Plan Mode Issues
| Symptom | Fix |
|---|---|
| Plan Mode locked | Pro/Max required |
| Plan too large/vague | Smaller goal; split into two plans |
| Step failed mid-run | Re-run plan; edit steps before execute |
| Credits exhausted mid-plan | Top up Billing; resume or restart |
Plan Mode replaces deprecated Classic Pipelines for multi-step work.
Legacy Pipelines
Classic Pipelines are deprecated (sunset July 20, 2026). They are hidden unless enabled under Profile → Customize → Enable Pipelines experiment.
Instead use: Plan Mode for dynamic multi-step goals; Recipes for repeatable single prompts.
Schedules and Automations
| Symptom | Fix |
|---|---|
| Can't create schedule | 10 active schedules max — deactivate old ones |
| Schedule disabled | Paid plan; enterprise automations policy |
| Post-meeting rule didn't fire | Check meeting matched filters; automation enabled |
Manage under Automations.
Credits and Billing
| Message | Meaning |
|---|---|
| Insufficient credits | Top up in Settings → Billing |
| Feature requires Pro | Upgrade plan |
| NO_CREDITS in Canvas SDK | Same — artifact AI calls use credits |
Agent and image steps consume more credits than simple chat.
Canvas and Artifacts
| Symptom | Fix |
|---|---|
| Canvas didn't open | Say "open in canvas" or Open in Canvas |
| Storage not saving | Publish artifact first (springbase.storage) |
| AI SDK errors in artifact | HTML artifacts only; check credits |
See Artifacts & Canvas.
Enterprise-Specific
- restrict_public_sharing — use Enterprise/Team visibility instead of public links.
- SSO enforce — password sessions may be revoked; sign in via SSO.
- Retention policies — old chats/docs may be deleted on schedule — export if needed.
See Security and Sharing.
Still Stuck?
- Note which surface (Chat, Recipe, Meeting, Agent, Plan, Canvas).
- Capture the exact error message or screenshot.
- Check Billing (plan + credits).
- Contact support with Context/Recipe/meeting ID if applicable.
Quick Links
- Quickstart — first setup path
- Key Concepts — terminology
- Overview — product map
Related docs
Learn how to use Springbase Chat with Contexts, @mentions, shared knowledge, live sources, saved outputs, permissions, and practical prompt templates.
Learn how to build living knowledge bases in Springbase with documents, notes, shared access, live sources, community Contexts, and grounded Chat.
Keep Contexts fresh with scheduled website crawls, RSS feeds, YouTube videos, and connected cloud folders — without re-uploading files every week.
Turn reliable prompts into reusable templates with variables, optional Contexts, scheduling, sharing, and Agent mode for connected apps.
Run AI that reads and acts in your connected apps — Gmail, Slack, Calendar, GitHub, Notion, and 100+ tools — with bounded instructions and credit-based billing.
Record calls, join Zoom/Teams/Meet links, upload audio, or paste transcripts — then get summaries, action items, search, and Chat integration.
Get from first chat to a grounded Springbase workflow in minutes.