Skip to content

Config reference: keywords

  • Enterprise tuning surface

    Defaults + constraints are rendered directly from Pydantic.

  • Env keys when available

    Many fields have an env-style alias (from TriBridConfig.to_flat_dict()).

  • Tooltip-level guidance

    If a matching glossary entry exists, you’ll see deeper tuning notes.

Config reference Config API & workflow Glossary

Total parameters: 5

Group index
  • (root)

(root)

JSON key Env key(s) Type Default Constraints Summary
keywords.keywords_auto_generate KEYWORDS_AUTO_GENERATE int 1 ≥ 0, ≤ 1 Auto-generate keywords
keywords.keywords_boost KEYWORDS_BOOST float 1.3 ≥ 1.0, ≤ 3.0 Score boost for keyword matches
keywords.keywords_max_per_repo KEYWORDS_MAX_PER_REPO int 50 ≥ 10, ≤ 500 Max discriminative keywords per repo
keywords.keywords_min_freq KEYWORDS_MIN_FREQ int 3 ≥ 1, ≤ 10 Min frequency for keyword
keywords.keywords_refresh_hours KEYWORDS_REFRESH_HOURS int 24 ≥ 1, ≤ 168 Hours between keyword refresh

Details (glossary)

keywords.keywords_auto_generate (KEYWORDS_AUTO_GENERATE) — Auto-Generate Keywords

Category: general

Automatically extract repository keywords from code and documentation during indexing (1=yes, 0=no). When enabled, the system analyzes class names, function names, docstrings, and comments to build a keyword set for routing. This supplements manually-defined keywords in repos.json. Auto-generation is useful for new repos or when you don't know what routing keywords to use. Disable if you prefer full manual control via repos.json.

Recommended: 1 for automatic keyword discovery, 0 for strict manual control.

Badges: - Multi-repo feature - Complements manual keywords

keywords.keywords_boost (KEYWORDS_BOOST) — Keywords Boost

Category: general

Score boost multiplier applied to search results that match corpus keywords. Higher values (1.5-2.0) strongly favor keyword matches, lower values (1.1-1.3) provide mild preference. The boost is multiplied with the base retrieval score. Recommended: 1.3 for balanced keyword preference.

Sweet spot: 1.3 for balanced systems. Use 1.1-1.2 for mild keyword preference (keyword matches slightly favored). Use 1.5-2.0 when keywords are highly reliable indicators of relevance. Use 2.5+ only when keywords are definitive relevance signals.

• Range: 1.0-3.0 (typical: 1.1-2.0) • Mild boost: 1.1-1.2 (slight preference) • Balanced: 1.3 (recommended) • Strong boost: 1.5-2.0 (strong preference) • Effect: Higher = more weight to keyword matches • Symptom too low: Keyword matches undervalued • Symptom too high: Keyword matches dominate, other signals ignored

Badges: - Scoring

Links: - Score Boosting - TF-IDF Scoring - Keyword Extraction

keywords.keywords_max_per_repo (KEYWORDS_MAX_PER_REPO) — Keywords Max Per Repo

Category: general

Maximum number of repository-specific keywords to extract and store for query routing in multi-repo setups. Higher values (100-200) capture more routing signals but increase memory and may introduce noise. Lower values (20-50) keep routing focused on core concepts. Keywords are extracted from code, docs, and enrichment metadata. Used by the router to determine which repositories are most relevant for a given query.

Recommended: 50-100 for most repos, 150-200 for large multi-domain codebases, 20-30 for focused microservices.

Badges: - Multi-repo only - Auto-generated

keywords.keywords_min_freq (KEYWORDS_MIN_FREQ) — Keywords Min Frequency

Category: general

Minimum term frequency required for a keyword to be included. Terms must appear at least this many times in the corpus to be considered. Higher values (5-10) ensure keywords are common enough to be meaningful, lower values (1-3) allow rare but distinctive terms. Recommended: 3 for balanced filtering.

Sweet spot: 3 for most corpora. Use 1-2 when you want to include rare but distinctive terms (e.g., unique function names). Use 5-7 when you want only common, well-established keywords. Use 10+ only for very large corpora.

• Range: 1-10 (typical: 2-5) • Rare terms: 1-2 (include distinctive rare terms) • Balanced: 3 (recommended) • Common terms: 5-7 (only well-established keywords) • Effect: Higher = fewer keywords, more common terms only • Symptom too low: Rare, potentially noisy keywords included • Symptom too high: Important distinctive keywords filtered out

Badges: - Keyword Extraction

Links: - TF-IDF Keyword Extraction - Term Frequency - Keyword Extraction

keywords.keywords_refresh_hours (KEYWORDS_REFRESH_HOURS) — Keywords Refresh (Hours)

Category: general

How often (in hours) to regenerate repository keywords from code for improved query routing. Lower values keep keywords fresh but increase indexing overhead. Typical: 24-168 hours (1-7 days).

Links: - Keyword Extraction