feat(models): add Kimi/Moonshot as overflow fallback#13
Merged
jonathanpeterwu merged 9 commits intomainfrom May 3, 2026
Merged
feat(models): add Kimi/Moonshot as overflow fallback#13jonathanpeterwu merged 9 commits intomainfrom
jonathanpeterwu merged 9 commits intomainfrom
Conversation
added 9 commits
April 29, 2026 09:45
…a exhausted Adds moonshot (Kimi K2.6) as a provider throughout the routing stack. When Claude CLI or API hits rate limits/quota, tasks automatically overflow to Kimi at ~10x lower cost ($0.60/$2.50 per MTok). Sensitive content stays on Anthropic via existing guard.
Analyzes stored traces to detect repeated failure patterns (lint, test, build, timeout, rate-limit), verification gaps, retry loops, and context thrash. Generates actionable recommendations with confidence scores and persists reports to .stackmemory/build/.
- Wire ContentCache into MCP server tool dispatch (28 cacheable read-only tools) - Add lookupByKey/putByKey for input-addressed caching (tool+args → result) - Add cache_stats + cache_lookup MCP tools for in-session token savings visibility - Print cache hit rate + tokens saved to stderr on MCP server exit - Add `stackmemory pack` CLI (install/list/search/show/init/fork/publish/uninstall) - Support local dirs, pack.yaml paths, GitHub URLs, and namespace/name shorthand - Create 3 first-party skill packs: coding/typescript-react, coding/python-fastapi, ops/decision-recovery - 6 new tests for key-based cache operations, all 2172 tests passing
…tools Three new core modules for stackmemory Q1: - src/core/provenance/ — TraceEvent spec (ASI-shaped), confidence scorer (ported from provenantai), provenance store with supersession + lineage - src/core/skill-packs/ — pack.yaml Zod schema, YAML parser, SQLite registry with FTS5 search, namespace/runtime filtering - src/core/cache/ — Content-hash token cache with SHA-256 dedup, hit counting, eviction, and savings stats MCP server wired with 7 new tools: cache_lookup, cache_stats, pack_list, pack_search, pack_get, record_trace, score_confidence. 102 tests passing across all modules.
Self-contained SDK at packages/sdk/ with typed facade over:
- ContentCache (SHA-256 dedup, token savings tracking)
- SkillPackRegistry (install/search/list, FTS5, pack.yaml parser)
- ProvenanceStore (TraceEvent spec, lineage, supersession)
- scoreConfidence() (decision detection, weighted signals)
Usage: new StackMemory({ dataDir }) → sm.cache / sm.packs / sm.provenance
16 tests passing. Zero type errors. Ready for npm publish.
- Add ASI-shaped TraceEvent type matching kickoff spec (score, feedback, provenance, cost, tokens) - Add TraceEventStore with SQLite persistence, filtered queries, batch recording, annotation - Add provenance columns to frames + anchors tables (source, derivation, confidence, superseded_by) - Populate provenance automatically on frame/anchor creation - Add `stackmemory cache stats/clear/search` CLI for terminal-printable token savings - 19 new tests for trace event store, all 2191 tests passing
- Wire TraceEventStore into MCP server with query/stats/record handlers - Add verification commands to multimodal harness (custom pass/fail checks) - Deterministic critique now checks verification results - Update MCP tool definitions + docs
SDK ready for npm publish at 17.9 kB.
- Add @StackMemory Python SDK (packages/python-sdk/) with cache, packs, provenance - Zero external deps (stdlib sqlite3), 12 tests passing - Wire TraceEventStore into MCP server — every tool call recorded as ASI-shaped event - Add trace_events, trace_event_stats, trace_event_annotate MCP tools - Add stackmemory cache stats/clear/search CLI commands
This file contains hidden or bidirectional Unicode text that may be interpreted or compiled differently than what appears below. To review, open the file in an editor that reveals hidden Unicode characters.
Learn more about bidirectional Unicode characters
Sign up for free
to join this conversation on GitHub.
Already have an account?
Sign in to comment
1 participant
Add this suggestion to a batch that can be applied as a single commit.This suggestion is invalid because no changes were made to the code.Suggestions cannot be applied while the pull request is closed.Suggestions cannot be applied while viewing a subset of changes.Only one suggestion per line can be applied in a batch.Add this suggestion to a batch that can be applied as a single commit.Applying suggestions on deleted lines is not supported.You must change the existing code in this line in order to create a valid suggestion.Outdated suggestions cannot be applied.This suggestion has been applied or marked resolved.Suggestions cannot be applied from pending reviews.Suggestions cannot be applied on multi-line comments.Suggestions cannot be applied while the pull request is queued to merge.Suggestion cannot be applied right now. Please check back later.
Summary
sensitive-guard— moonshot is intentionally NOT inAPPROVED_PROVIDERSFALLBACK_CHAINandCHEAP_PROVIDERSfor cost-optimal routingTest plan