viktor/beadboard
1
0
Fork 0
Code Issues Pull requests Projects Releases Packages Wiki Activity Actions

Update NEXT_SESSION_PROMPT.md for beadboard-05a epic

Browse source
This commit is contained in:
zenchantlive 2026-03-02 19:33:45 -08:00
parent 64a5129412
commit 9010c6a058
1 changed files with 143 additions and 97 deletions
Download patch file Download diff file Expand all files Collapse all files

240
NEXT_SESSION_PROMPT.md
View file

@ -1,113 +1,159 @@
# Next Session: Investigate Frontend/Dolt Data Mismatch
# Next Session: Power Through [EPIC] Global BeadBoard Installer + Runtime Launch (`beadboard-05a`)
## Critical Unresolved Issue
## Mission
The frontend is showing STALE data that doesn't match Dolt. We could NOT figure out the root cause.
Complete `beadboard-05a` end-to-end with strict TDD and evidence-first closeout.
### Symptoms:
1. **bb-u6f.7** shows as OPEN in frontend left sidebar, but is CLOSED in Dolt
2. **beadboard-2e6** and **beadboard-6cc** (brainstorming epics) don't appear in frontend, but exist in Dolt as OPEN
3. Memory-anchor epics ARE correctly filtered out (this works)
4. Frontend loads without Dolt connection errors
This epic implements a single cross-platform BeadBoard installer system (Windows/Linux/macOS wrappers over one shared contract), global `bb`/`beadboard` command availability, runtime launcher behavior, and driver-skill detection-only remediation messaging.
### What We Verified:
- Dolt IS running (bd dolt status shows PID 77304, port 3307)
- Dolt has CORRECT data (direct MySQL queries confirmed bb-u6f.7=status:closed, beadboard-2e6=status:open)
- issues.jsonl has WRONG data (bb-u6f.7=status:open, issue_type=task - both wrong!)
- No errors in console or Dolt logs
## Non-Negotiable Context
### What We Tried (ALL FAILED):
1. Added memory-anchor filter to left-panel.tsx line 73 - worked for filtering memory, but didn't fix stale data
2. Removed issues.jsonl fallback in read-issues.ts - didn't change anything
3. Cleared .next cache and restarted dev server - didn't change anything
4. Restarted Dolt server - didn't change anything
5. Multiple browser hard refreshes - didn't change anything
1. BeadBoard global install is for humans/operators.
2. Driver skill does not install BeadBoard; it only checks install/discovery and proceeds.
3. Bead hierarchy in this epic must stay `beadboard-05a.x(.x)` style.
4. Use `bd` as source of truth; no direct JSONL writes.
5. Evidence before completion claims.
### What I Thought Was The Cause (WRONG):
1. ~~Dolt server not running~~ - it WAS running
2. ~~Frontend falling back to issues.jsonl~~ - removed fallback, same result
3. ~~Next.js build cache~~ - cleared, same result
4. ~~LeftPanel filter hiding epics~~ - verified filter logic is correct
5. ~~Dolt connection failing silently~~ - but no error means Dolt IS connecting
## First 10 Minutes (Start From Scratch)
### Files Modified This Session:
- `src/components/shared/left-panel.tsx` - added memory-anchor filter (line 73)
- `src/lib/read-issues.ts` - removed issues.jsonl fallback (Dolt-only now)
### Files To Read To Understand What Happened:
1. `src/lib/read-issues.ts` - Dolt-only read path
2. `src/lib/read-issues-dolt.ts` - Dolt query logic
3. `src/lib/aggregate-read.ts` - calls readIssuesFromDisk
4. `src/app/page.tsx` - SSR entry point
5. `src/components/shared/left-panel.tsx` - epic display logic
6. `.beads/dolt-server.log` - connection logs (no errors found)
### Key Questions For Next Session:
1. Why does the frontend show different data than Dolt when Dolt is clearly accessible?
2. Is there another caching layer we're missing?
3. Is there something in the Next.js SSR that's caching data?
4. Could there be a second Dolt instance or database being read?
---
## What Was Accomplished This Session:
### Completed Earlier:
1. Fixed BlockedTriageModal CSS theme variables
2. Fixed agent names displaying as bead IDs (extract human-readable names)
3. Fixed blocked count to include derived blockers
4. Created memory node beadboard-6iq (Extract human-readable names)
5. Closed old bb-* epics (they were from old architecture)
6. Created beadboard-1bg epic for skill v4 rewrite (closed - all tasks done)
7. Created beadboard-n1h epic for quality gates
8. Created beadboard-jq5 epic for project scope UI (closed)
9. Created beadboard-2e6 epic for UX critique brainstorming
### Memory Nodes Created:
- beadboard-6iq - [MEMORY][UI][HARD] Extract human-readable names from raw data fields
- beadboard-dc0 - [MEMORY][ARCH][SOFT] Skill structure pattern
### Skills Used This Session:
1. `systematic-debugging` - followed the debugging process but couldn't find root cause
2. `beadboard-driver` - used bd commands throughout
3. `verification-before-completion` - ran gates before closing beads
---
## Ready Work (from bd ready):
- `beadboard-n1h.1` - Unit Tests: Core Libraries
- `beadboard-n1h.2` - API Integration Tests
- `beadboard-2e6` - [BRAINSTORM] UX Continuity and Critique (needs brainstorming skill)
- Other quality/testing tasks
---
## Non-Negotiable Context:
1. Run `bd` commands from `codex/beadboard` directory
2. `bd` is source of truth - no direct JSONL writes
3. Evidence before assertions - always cite command output
4. Read hard memory first: `bd show beadboard-116 beadboard-60a beadboard-zas`
## First 10 Minutes:
Run exactly:
```bash
cd /mnt/c/Users/Zenchant/codex/beadboard
# 1) Read hard memory
bd show beadboard-116 beadboard-60a beadboard-zas beadboard-6fv
# 1) Read hard memory / contract memory
bd show beadboard-116 beadboard-60a beadboard-zas beadboard-6fv beadboard-at4
# 2) Check current state
bd query "status=closed" --sort closed --reverse --limit 10
bd ready
# 2) Load recent context
bd query "status=closed" --sort closed --reverse --limit 20
bd ready -n 25
# 3) Check Dolt status
bd dolt status
# 3) Inspect epic + task tree
bd show beadboard-c70 beadboard-05a
bd dep tree beadboard-05a
# 4) Verify data consistency
# Compare Dolt vs frontend for a known-closed epic
bd show bb-u6f.7
# Then check what frontend shows
# 4) Create your session agent bead and claim work
bd create --title="Agent: <role-name>" --description="Own beadboard-05a execution" --type=task --priority=0 --label="gt:agent,role:orchestrator"
bd update beadboard-05a --status in_progress --assignee <your-agent-bead-id>
```
## Required Skills To Use
Use these skills explicitly during implementation:
1. `using-superpowers`
2. `beadboard-driver`
3. `test-driven-development`
4. `verification-before-completion`
5. `linus-beads-discipline`
6. `writing-plans` (if you need to re-slice tasks before coding)
## Latest Closed Beads (Most Relevant)
Recent close context that matters here:
- `beadboard-x1y` - Hide Closed regression fix (epic visibility bug)
- `beadboard-om4` - [MEMORY] Stale UI parity triage order
- `beadboard-cyk` - [MEMORY] HideClosed invariants
- `beadboard-btt` - Driver install check contract + remediation messaging
- `beadboard-i0q` - Global skill project-context contract
Current open execution epic:
- `beadboard-05a` - `[EPIC] Global BeadBoard Installer + Runtime Launch`
## Exact Task IDs For Execution
Work this exact tree:
- `beadboard-05a.1` Installer Contract: canonical manifest + shared semantics
- `beadboard-05a.1.1` Installer Contract ADR
- `beadboard-05a.1.2` Manifest Schema + Validation
- `beadboard-05a.2` Windows installer wrapper
- `beadboard-05a.2.1` Windows one-liner + PATH contract
- `beadboard-05a.3` beadboard launcher (start/open/status)
- `beadboard-05a.4` Linux/mac installer wrapper
- `beadboard-05a.5` Driver detection alignment
- `beadboard-05a.6` Installer CI + smoke tests
- `beadboard-05a.7` Installer docs + operator quickstart
Execution order is dependency-driven; do not bypass.
## Important Files To Read First
Read these before coding:
1. `/mnt/c/Users/Zenchant/codex/beadboard/AGENTS.md`
2. `/mnt/c/Users/Zenchant/codex/beadboard/skills/beadboard-driver/SKILL.md`
3. `/mnt/c/Users/Zenchant/codex/beadboard/skills/beadboard-driver/scripts/session-preflight.mjs`
4. `/mnt/c/Users/Zenchant/codex/beadboard/skills/beadboard-driver/scripts/diagnose-env.mjs`
5. `/mnt/c/Users/Zenchant/codex/beadboard/skills/beadboard-driver/scripts/resolve-bb.mjs`
6. `/mnt/c/Users/Zenchant/codex/beadboard/skills/beadboard-driver/scripts/lib/driver-lib.mjs`
7. `/mnt/c/Users/Zenchant/codex/beadboard/tools/bb.ts`
8. `/mnt/c/Users/Zenchant/codex/beadboard/scripts/bb-init.mjs`
9. `/mnt/c/Users/Zenchant/codex/beadboard/package.json`
10. `/mnt/c/Users/Zenchant/codex/beadboard/docs/adr/2026-02-14-beadboard-driver-skill-and-bb-resolution.md`
Also note current state:
- `install/` directory does not yet exist; this epic will introduce it.
## Strict TDD Execution Flow (Per Task)
For every `beadboard-05a.x` task:
1. `bd update <task-id> --status in_progress --assignee <your-agent-bead-id>`
2. Write failing test(s) first.
3. Run focused test and confirm red for correct reason.
4. Implement minimum code to pass.
5. Re-run focused tests (green).
6. Run broader relevant suite.
7. Record evidence in notes:
- commands run
- pass/fail output summary
8. Close only after evidence is recorded.
Do not claim complete without fresh command output.
## Verification Gates
Before closing each code-changing bead and before epic close:
```bash
npm run typecheck
npm run lint
npm run test
```
If non-epic unrelated failures appear, cite exact failing file/test in bead notes and proceed transparently.
## Implementation Guardrails
1. Keep installer architecture unified: one canonical manifest contract + thin OS wrappers.
2. Keep skill boundary strict: detection/remediation only, no install side effects.
3. Preserve Windows support as mandatory while implementing Linux/mac parity.
4. Keep user-facing install/docs copy simple and explicit.
5. Do not resurrect deferred legacy tasks (`beadboard-ydu`, `beadboard-27u`, etc.); use `beadboard-05a.x` tree only.
## Expected Deliverables At Epic Close
1. New installer contract + manifest definition.
2. `install.ps1` and `install.sh` wrappers aligned to the same manifest semantics.
3. Global `bb` + `beadboard` command behavior implemented.
4. `beadboard` launcher behavior: start/open/status contract.
5. Driver detection messaging updated for platform install remediation.
6. CI/smoke tests for install/reinstall/failure modes.
7. Operator docs with one-liners and runtime command behavior.
8. `bd` notes/close reasons with full verification evidence.
## Session Landing Checklist
Before handing off:
1. `bd ready` and confirm unblocked next steps are accurate.
2. Update `NEXT_SESSION_PROMPT.md` with what changed, verified commands, risks, and exact next task.
3. Memory review:
- if reusable new rule: create/supersede canonical memory bead
- else add note: "Memory review: no new reusable memory."
4. Close your session agent bead.