atomic

flowstate-epic-flowstate-community-epicdm-flowstate-rag-client

Package-local agent skill for @epicdm/flowstate-rag-client.

Tags

Publisher

Versions

1 version published — latest: 1.0.7

  • 1.0.7
    publishedInvalid Date

Skill

flowstate-rag-client Package Skill

Package: @epicdm/flowstate-rag-client Repository: epic-flowstate-community Path: packages/flowstate-rag-client Version: 1.0.7

Use this published Dojo skill as the package-level entrypoint. For detailed workflows, use the package-local skills listed below.

Primary Package Workflow

RAG Client Codebase Search

Status: Active Purpose: Use metadata filters for code and codebase search correctly. Scope: MetadataFilters, codebaseId, code collection queries, MCP codebase wrappers. Trigger: Codebase semantic search, code context assembly, MCP codebase-search behavior review. Input: Connected RAGClient, query, workspace/codebase metadata. Output: Search results scoped to indexed code or codebase documents.


Workflow

Use runtime setup first. Then pass metadata/codebase filters rather than relying on broad document search.

Check MCP wrapper mappings in:

  • packages/flowstate-mcp/src/tools/CodebaseSearchTools.ts
  • packages/flowstate-mcp/src/tools/RAGTools.ts

Pitfalls

  • Codebase indexing schema is shared with document-store/rag-sync.
  • This package validates some metadata/codebase inputs, but not every option.
  • Root README omits some codebase filter behavior; source and tests are authority.

Done When

  • The query is scoped to the intended codebase/workspace.
  • Upstream indexed fields match RAGClient expectations.
  • MCP wrapper defaults are checked when using MCP tools.

Package Skill Inventory

Package Skill Inventory: @epicdm/flowstate-rag-client

Package Context

Package path: /Users/sthornock/code/epic/epic-flowstate-community/packages/flowstate-rag-client

This package is a local RAG query client. It should teach agents how to configure runtime dependencies and use the query workflows, not just list RAGClient.

Skill Map

SkillTypeTriggerTeachesEvidenceDepends OnDojo
flowstate-rag-clientorchestratorUse when integrating or changing @epicdm/flowstate-rag-clientRoutes setup, semantic context, codebase search, memory/stats, troubleshootingsrc/index.ts, src/RAGClient.tsrag-sync/document-store/mcp contextready-after-review
flowstate-rag-client-runtime-setupreferenceUse when configuring SurrealDB/Ollama for local RAG queriesConfig fields, env mapping, model parity, rag_documents schemaRAGClientConfig, sync schema, testsRAG sync setupready-after-review
flowstate-rag-client-semantic-contextworkflowUse when performing search or building LLM contextsearch(), getContext(), ranking, filters, token limits, defaultssearch, getContext, testsruntime setupready-after-review
flowstate-rag-client-codebase-searchworkflowUse when querying indexed codecodebaseId, MetadataFilters, code collection, MCP wrapper mappingMetadataFilters, CodebaseSearchToolsdocument-store code indexerready-after-review
flowstate-rag-client-memory-similarity-statsworkflowUse when recalling memories, finding similar docs, or checking index statsrecall(), findSimilar(), getStats(), namespace/collection rulesRAGClient.ts, testsruntime setupready-after-review
flowstate-rag-client-troubleshooting-maintenancetroubleshootingUse when RAG queries fail or docs/contracts changeNot connected, Ollama embeddings, SurrealDB errors, empty index, downstream drifttests, MCP wrappers, smoke driftall abovelocal-only

Anti-Sprawl Decision

Six skills are appropriate: one orchestrator, one setup/schema reference, three recurring workflows, and one troubleshooting/maintenance skill. Per-method skilllets would hide shared runtime and schema constraints.

Feature Coverage

Feature GroupSource EvidenceCovered ByNotes
Public client/exportsrc/index.tsorchestratorSingle entrypoint
Runtime setupRAGClientConfig, connect()runtime setupSurrealDB/Ollama required
Semantic searchsearch()semantic-contextMemories excluded by default
Context buildinggetContext()semantic-contextRough token estimate
Codebase filtersMetadataFilterscodebase-searchUsed by MCP wrappers
Memory recallrecall()memory-similarity-statsNamespace maps to org_id
Similarity lookupfindSimilar()memory-similarity-statsSame collection
StatsgetStats()memory-similarity-statsMissing from root README

Composition Model

Start with the orchestrator. Load runtime setup for services/config. Choose semantic-context, codebase-search, or memory-similarity-stats by task. Use troubleshooting before changing contracts shared with MCP, rag-sync, or document-store.

Deferred Skills

  • SurrealDB schema authoring belongs to sync/document-store until this package owns schema writes.
  • Ollama embedding operations are limited to model parity and failure handling here.
  • Dojo publish remains deferred until generated content is rebuilt.

Review Questions

  • Should zod remain a dependency or should source add schemas?
  • Should root README/changelog be replaced by .flowstate/docs content?
  • Should RAGClient.search() validate query/limit/minScore/collections?
  • Should document-store smoke insert canonical content/doc_id fields?

Review Gate

Package Skills Review: @epicdm/flowstate-rag-client

Verdict

Pass with fixes.

The local skillset now covers the real RAG client workflows: runtime setup, semantic context, codebase search, memory/similarity/stats, and troubleshooting. Existing generated Dojo content remains unsuitable for publication.

Findings

SeveritySkillIssueEvidenceRequired Fix
ImportantallExisting Dojo is generated from one coarse feature.flowstate/dojo/*Rebuild from reviewed skills
Importantruntime/troubleshootingSchema ownership spans rag-sync/document-storeSource and downstream refsKeep contract warnings in setup/troubleshooting
MinortroubleshootingVerification passed after authoring; keep this evidence current before Dojo publishtest, typecheck, build passed in this passRe-run before final Dojo publish if source changes
Minorruntimezod dependency not used in sourcepackage/sourceTrack maintenance question

Coverage

Feature GroupStatusSkillNotes
Runtime setupCoveredruntime setupSurrealDB/Ollama/model parity
Semantic search/contextCoveredsemantic-contextIncludes token estimate
Codebase filteringCoveredcodebase-searchIncludes MCP wrappers
Memory recall/similarity/statsCoveredmemory-similarity-statsIncludes namespace semantics
Troubleshooting/maintenanceCoveredtroubleshooting-maintenanceIncludes wrapper/default drift

Drift and Contract Risks

AreaRiskEvidenceDecision
DojoCoarse generated feature course.flowstate/dojo/*Rebuild before publish
README/changelogMissing newer APIs/version driftREADME/CHANGELOG vs sourceSource is authority
Document-store smokeNoncanonical fieldssmoke script vs client selectsTrack cross-package follow-up
MCP wrappersDefaults differ from raw clientRAGTools vs search()Teach wrapper distinction

Dojo Publication Decision

Do not publish existing generated Dojo files. Reviewed local skills can become Dojo source after verification and manifest rebuild.

Follow-Up Work

  • Package verification passed:
    • yarn workspace @epicdm/flowstate-rag-client test
    • yarn workspace @epicdm/flowstate-rag-client typecheck
    • yarn workspace @epicdm/flowstate-rag-client build
  • Rebuild .flowstate/dojo artifacts.
  • Decide whether to add Zod validation or remove dependency.
  • Align README/changelog with source.

Usage

Start with this skill to understand the package purpose, then use the package-local focused skills for specific workflows. Keep this Dojo version aligned with packages/flowstate-rag-client/package.json. Do not treat generated feature lists as lessons; use workflow evidence from CODE-REVIEW.md, SKILL-INVENTORY.md, and REVIEW.md.

Install

Run this in your project to install via CLI:

fscloud dojo skill install flowstate-epic-flowstate-community-epicdm-flowstate-rag-client --target <target> --out ./skills

Replace <target> with your agent target identifier. See the CLI docs for details.