Aim Github Search

Search GitHub - Semantic Search for GitHub Content

Search the github collection for GitHub-sourced content using semantic similarity with advanced filtering.

Activation

# Basic semantic search
/aim-github-search "authentication bug"

# Filter by type
/aim-github-search "API refactoring" --type github_pr
/aim-github-search "CI failures" --type github_ci_result

# Filter by state
/aim-github-search "deployment fix" --state merged

# Combine filters
/aim-github-search "security" --type github_issue --limit 10

Options

• --type <type> - Filter by GitHub document type: github_issue, github_pr, github_commit, github_ci_result, github_code_blob, github_issue_comment, github_pr_review, github_pr_diff
• --state <state> - Filter by state: open, closed, merged
• --limit <n> - Maximum results to return (default: 5)

Result Format

Each result includes:

• GitHub URL - Direct link
• Metadata badges - Type, State, Date
• Content snippet - First ~300 characters
• Relevance score - Semantic similarity with decay (0-100%)

Qdrant Connection Details

GitHub content is stored in the github collection:

Qdrant Payload Schema

Every GitHub point in github has the following payload fields. Use these exact names for filtering.

Common Fields (all GitHub points)

Direct Query Examples

Use query.py for direct Qdrant queries. The script applies source=github automatically and accepts optional --type, --state, --limit, and --format flags.

Search by source

# Replace "owner/repo-name" with your GITHUB_REPO value (e.g., "hidden-history/ai-memory")
# --collection defaults to github; omit it for the standard case
bash "${AI_MEMORY_INSTALL_DIR:-$HOME/.ai-memory}/scripts/memory/run-with-env.sh" \
  "${AI_MEMORY_INSTALL_DIR:-$HOME/.ai-memory}/_ai-memory/skills/aim-github-search/scripts/query.py" \
  --group-id "owner/repo-name"

Filter by type and state

# Replace "owner/repo-name" with your GITHUB_REPO value
bash "${AI_MEMORY_INSTALL_DIR:-$HOME/.ai-memory}/scripts/memory/run-with-env.sh" \
  "${AI_MEMORY_INSTALL_DIR:-$HOME/.ai-memory}/_ai-memory/skills/aim-github-search/scripts/query.py" \
  --group-id "owner/repo-name" --type github_pr --state merged --limit 20

Count GitHub points in collection

# Returns the exact count via the Qdrant count endpoint (exact=true)
bash "${AI_MEMORY_INSTALL_DIR:-$HOME/.ai-memory}/scripts/memory/run-with-env.sh" \
  "${AI_MEMORY_INSTALL_DIR:-$HOME/.ai-memory}/_ai-memory/skills/aim-github-search/scripts/query.py" \
  --group-id "owner/repo-name" --format count

Query Script Reference

The parameterized query script at _ai-memory/skills/aim-github-search/scripts/query.py implements the source=github + group_id filter pattern. It accepts:

python3 query.py \
  --group-id GROUP_ID          # required; e.g. "hidden-history/ai-memory"
  [--type TYPE]                # github_issue | github_pr | github_commit |
                               #   github_ci_result | github_code_blob |
                               #   github_issue_comment | github_pr_review | github_pr_diff
  [--state STATE]              # open | closed | merged
  [--limit N]                  # default: 10
  [--format table|json|count]  # default: table; count uses Qdrant count endpoint (exact=true)
  [--collection COLL]          # default: github; use to query a different collection

Run via run-with-env.sh so memory.* imports and QDRANT_API_KEY are resolved automatically:

bash "${AI_MEMORY_INSTALL_DIR:-$HOME/.ai-memory}/scripts/memory/run-with-env.sh" \
  "${AI_MEMORY_INSTALL_DIR:-$HOME/.ai-memory}/_ai-memory/skills/aim-github-search/scripts/query.py" \
  --group-id "hidden-history/ai-memory" --type github_pr --state merged

Technical Details

• Semantic Search: Uses jina-embeddings-v2-base-en for vector similarity
• Tenant Isolation: Mandatory group_id filter prevents cross-repo leakage
• Performance: < 2s for typical searches
• Collection: github (dedicated collection for GitHub-sourced content, per PLAN-010)
• Score Threshold: Configurable via SIMILARITY_THRESHOLD (default 0.7)
• Port: 26350 (NOT the Qdrant default of 6333)
• API Key: Required -- stored in ~/.ai-memory/docker/.env as QDRANT_API_KEY
• Decay Scoring: Applied to results via existing search path

Notes

• GitHub repo is auto-detected from project configuration
• Results sorted by relevance score (highest first)
• All filters are optional except the search query
• Use exact field names from the schema above -- source NOT namespace, type NOT doc_type
• The source="github" filter is always applied to restrict results to GitHub content
• The group_id filter is always applied for mandatory tenant isolation (prevents cross-repo leakage)