Skip to content

Repository files navigation

Kiro

Solr Compass

Solr-backed semantic search, codebase retrieval, and cross-session memory for Kiro through a guided MCP workflow.

Release Version GitHub Packages License Kiro Power

Bun Apache Solr Docker MCP

Features

  • Vector, BM25 keyword, and hybrid search across multiple Solr collections
  • Codebase indexing with automatic chunking and incremental sync
  • User document storage and retrieval
  • Durable cross-session memory notes
  • Workspace profiling against knowledge artifacts
  • Local embedding by default (no cloud credentials required) or optional Amazon Bedrock Titan provider
  • Context-aware content resolution with relevance feedback loops
  • Optional Zotero integration for formal citation export

Installation

From GitHub

  1. Open Kiro's Powers panel
  2. Click Add power from GitHub
  3. Paste this repository URL:
    https://github.com/stevenjmiklovic/solrcompass
    
  4. Kiro installs the power and configures the MCP server automatically

From a Local Clone

git clone https://github.com/stevenjmiklovic/solrcompass.git

Then in Kiro: Powers panel → Add power from Local Path → select the cloned directory.

Prerequisites

  • Bun 1.0 or later on PATH
  • Docker Desktop for the bundled local SolrCloud environment

No additional installation steps are needed. Kiro launches the version-pinned npm package @stevenjmiklovic/solrcompass@0.2.0 through bunx --bun automatically.

Usage

Once installed, mention any of these keywords in Kiro chat to activate the power:

  • "semantic search", "solr", "compass"
  • "index codebase", "search code"
  • "remember this", "recall memory"

Or invoke it explicitly: ask Kiro to "use the Solr Compass power to index this repository."

On activation, the power:

  1. Checks that Solr is reachable via Docker
  2. If not running, offers to initialize a local SolrCloud instance (pulls solr:10.0.0 and zookeeper:3.9.3 images)
  3. Creates persistent collections for artifacts, documents, and source code
  4. Proceeds with your requested search or indexing task

Quick Start Examples

Index and search a codebase:

compass_index_folder { "path": "/path/to/your/repo", "chunked": true }

compass_search_codebase {
  "query": "Where is the authentication middleware?",
  "root": "/path/to/your/repo",
  "mode": "hybrid",
  "topK": 10
}

Store and recall a project note:

compass_remember {
  "note": "This project uses Bun for builds and tests.",
  "category": "convention",
  "tags": ["my-project"]
}

compass_recall_memory {
  "query": "How does this project run tests?",
  "tags": ["my-project"],
  "topK": 5
}

See the full POWER.md for complete workflow documentation, tool reference, configuration variables, and troubleshooting.

Skills

Solr Compass ships with six specialized skills that cover the complete library-retrieval workflow:

Skill Purpose
search-strategist Multi-pass search strategies, query optimization, mode selection, and answer synthesis
content-resolver Expand results to full content, context budgeting, source reconciliation, provenance tracking, and optional Zotero citation export
codebase-indexer Repository analysis, chunking strategy selection, indexing, and validation
memory-curator Organize, audit, deduplicate, and bootstrap cross-session memory notes
workspace-profiler Technology stack profiling, artifact matching, and knowledge gap analysis
collection-admin Health monitoring, collection lifecycle, provider migration, and remote SolrCloud configuration
Acquire → Catalog → Store → Discover → Retrieve & Apply → Maintain
(indexer)  (profiler) (admin) (strategist)  (resolver)     (curator + admin)

Steering Files

Step-by-step workflow guides loaded on demand:

File Purpose
semantic-search-workflow Guided search across scopes and modes
codebase-onboarding Step-by-step repository indexing with validation
memory-management Cross-session memory best practices
bedrock-embedding Amazon Bedrock Titan configuration and migration
troubleshooting Common failures, diagnostics, and recovery

Embedding Providers

Provider Model Dimensions Max Input Requirements
local (default) Xenova/all-MiniLM-L6-v2 384 padded to 1024 512 tokens None (runs on CPU)
bedrock-titan amazon.titan-embed-text-v2:0 1024 native 8,192 tokens Bedrock API key

See steering/bedrock-embedding.md for full Bedrock setup and migration instructions.

Repository Layout

solrcompass/
├── POWER.md                        # Power definition (loaded by Kiro on activation)
├── mcp.json                        # MCP server launch configuration
├── package.json                    # npm package published to GitHub Packages
├── steering/                       # Workflow guides
│   ├── semantic-search-workflow.md
│   ├── codebase-onboarding.md
│   ├── memory-management.md
│   ├── bedrock-embedding.md
│   └── troubleshooting.md
├── skills/                         # Specialized skill definitions
│   ├── search-strategist/SKILL.md
│   ├── content-resolver/SKILL.md
│   ├── codebase-indexer/SKILL.md
│   ├── memory-curator/SKILL.md
│   ├── workspace-profiler/SKILL.md
│   └── collection-admin/SKILL.md
├── .kiro/hooks/                    # Workspace hooks
│   ├── changelog-entry.json        # Auto-generate changelog on session end
│   └── incremental-index.json      # Auto-reindex on source file save
├── VERSION                         # Release version (triggers CI on change)
├── CHANGELOG.md                    # Release history
├── changes/                        # Unreleased changelog fragments
├── LICENSE                         # Apache-2.0
└── README.md

Origin

This power is sourced from the context-bazaar knowledge artifact catalog, compiled using the Kanon CLI. The canonical artifact lives at kanon/knowledge/solr-compass/.

License

Apache-2.0

About

A Kiro Power which utilizes Solr for local, fast vector search.

Topics

Resources

Stars

0 stars

Watchers

0 watching

Forks

Releases

Packages

Contributors