will/swarm-zap
1
0
Fork 0
Code Issues Pull Requests Actions Packages Projects Releases Wiki Activity
128 Commits 1 Branch 0 Tags
841365e0204d22c937e143f5f91f2dc6ba983293
Commit Graph

62 Commits

Author SHA1 Message Date
zap
23782735a1 docs(cost): update plan with corrections from official docs 
- Phase 1: clarify cacheRetention only applies to Claude models; GPT auto-caches; GLM has none
- Phase 1: add TTL reality check (short=5min, long=1h) and implications for heartbeat timing
- Phase 2: explain why long TTL + 25m heartbeat is the right combo
- Phase 4: replace generic prompt tips with model-specific guidance from official Anthropic/OpenAI docs
- Added prompt structure notes for cache efficiency, GLM-4.7 tighter prompting requirements
- References: memory/references/*.md
 2026-03-05 20:37:32 +00:00
zap
79e61f4528 docs(references): add Anthropic + OpenAI official best practices 
- anthropic-prompt-caching.md: KV cache mechanics, TTLs, pricing, auto vs explicit
- openai-prompt-caching.md: automatic caching, in-memory vs 24h retention, prompt_cache_key
- anthropic-prompting-best-practices.md: clear instructions, XML tags, few-shot, model-specific notes
- openai-prompting-best-practices.md: message roles, optimization framework, structured outputs, model selection

Key findings:
- Anthropic caching: only for Claude models, 5m default TTL, 1h optional, 10% cost for reads
- OpenAI caching: automatic/free, 5-10min default, 24h extended for GPT-5+
- GLM/ZAI models: neither caching mechanism applies
- Subagent model routing table added to openai-prompting-best-practices.md
 2026-03-05 20:34:38 +00:00
zap
c2fe8155e3 docs(cost): add upstream safety rules to cost optimization plan 
Hard constraints: no dist edits, no undocumented config keys, no middleware.
Safe: documented openclaw.json knobs + workspace files only.
Includes update workflow and pre-implementation verification checklist.
 2026-03-05 20:25:59 +00:00
zap
6642964ae6 docs(cost): add inference cost optimization plan — 4 phases 
Phase 1: Enable prompt caching (cacheRetention: long on Claude models)
Phase 2: Heartbeat cache warming (25m main, 55m default)
Phase 3: Context pruning (cache-ttl mode, 1h TTL)
Phase 4: Cheaper models for subagents (GLM-4.7 free tier for bulk work)

All config-only, no OpenClaw code changes, fully reversible.
 2026-03-05 20:20:03 +00:00
zap
c9fa2e1d95 docs(council): save D/P and personality run results + mode comparison 
- D/P run: 5 subagents, ~77k tokens, produced scored shortlists + merges
- Personality run: 4 subagents, ~62k tokens, produced narrative + verdict
- Comparison: D/P better for concrete ideas/scoring, personality better for adversarial tension/narrative
- Key finding: D/P lacks built-in skeptic, personality lacks structured scoring
- Proposed improvement: hybrid mode combining both strengths
 2026-03-05 19:44:34 +00:00
zap
0acd7a2772 chore(tasks): add Flynn council pipeline fix to TODO 
- medium priority, plan in memory/plans/flynn-council-fix.md
- 5 phases, estimated 1-2 focused sessions
 2026-03-05 19:01:49 +00:00
zap
e7051a617f docs(council): add Flynn council pipeline fix plan 
- 5-phase plan: config, structured output, bridge caps, E2E run, zap integration
- Work to happen on fix/council-pipeline branch in ~/flynn
- Goal: get Flynn's dual-council working so zap can delegate to it
 2026-03-05 19:00:58 +00:00
zap
3e198bcbb3 docs(council): add experimental findings from all 3 flow types 
- Tested parallel 1-round, sequential 1-round, debate/parallel 3-round
- 3 rounds is sweet spot: positions converge, meaningful evolution
- Sequential most token-efficient; parallel 3-round best depth-to-cost
- Debate and parallel 3-round mechanically identical (prompt tone differs)
- Added cost profiles, recommended defaults by use case
- Updated TODOs: unify flows, test 2-round, test mixed model tiers
 2026-03-05 16:39:32 +00:00
zap
da36000050 feat(council): add configurable rounds, flow parameters, and round-specific prompts 
- Parameters: flow (parallel/sequential/debate), rounds (1-5), tier (light/medium/heavy)
- Round-specific prompt templates: opening, rebuttal, final position
- Multi-round referee template tracks position evolution across rounds
- Word count guidance decreases per round to control token cost
- Subagent labeling convention: council-r{round}-{role}
- Updated from live testing with 1-round and 3-round parallel debates
 2026-03-05 16:21:22 +00:00
zap
7274d399ce feat(council): add council advisory skill with parallel/sequential/debate flows 
- 3 advisors (Pragmatist, Visionary, Skeptic) + 1 Referee subagent
- Default: parallel + synthesis flow
- Prompt templates in references/prompts.md
- Model tier selection based on topic complexity
- Added TODO tasks for personality depth and skill name revisit
 2026-03-05 08:41:35 +00:00
zap
b0b4060bf9 chore(workspace): install all 4 bundled hooks + model default update 
- Installed hooks: boot-md, session-memory, command-logger, bootstrap-extra-files
- Updated default model to litellm/copilot-claude-sonnet-4.6
- Added memory/startup-health.json + boot-last-run.json from boot check
- Added memory/startup-health.md warning log
- Captured new session memory files (2026-03-04/05)
- Minor MEMORY.md + USER.md: added git preference line
 2026-03-05 04:40:46 +00:00
zap
a8c73c2b84 chore(mcp): track smoke tool baseline 2026-03-05 02:18:32 +00:00