apunkt/vestige
1
0
Fork 0
mirror of https://github.com/samvallad33/vestige.git synced 2026-05-08 07:12:37 +02:00
Code Issues Projects Releases Packages Wiki Activity Actions

fix: push type filters into SQL WHERE clause + expose in MCP search

Browse source 
Type filtering (include_types/exclude_types) was applied post-fetch after
the database LIMIT, which could return zero results when all top-N
results were of the filtered type. This pushes type filters into the SQL
WHERE clause in keyword_search_with_scores() so they apply before the
limit. Semantic results still get post-fetch filtering as a safety net
since the vector index cannot filter by type.

Also adds hybrid_search_filtered() as the new primary method, with the
original hybrid_search() delegating to it with no filters for backward
compatibility. The MCP search tool now exposes include_types and
exclude_types parameters.

Includes 5 new test cases covering include, exclude, precedence,
empty results, and backward compatibility.

Co-Authored-By: Claude Opus 4.6 (1M context) <noreply@anthropic.com>
This commit is contained in:
Bot 2026-04-07 15:32:28 -05:00
parent 16fe2674ed
commit f3e25f7503
3 changed files with 565 additions and 314 deletions
Download patch file Download diff file Expand all files Collapse all files

23
crates/vestige-mcp/src/tools/search_unified.rs
View file

@ -66,6 +66,16 @@ pub fn schema() -> Value {
"items": { "type": "string" },
"description": "Optional topics for context-dependent retrieval boosting"
},
"exclude_types": {
"type": "array",
"items": { "type": "string" },
"description": "Node types to exclude from results (e.g., ['reflection']). Reflections are excluded by default to prevent polluting factual queries."
},
"include_types": {
"type": "array",
"items": { "type": "string" },
"description": "If set, only return nodes of these types. Overrides exclude_types."
},
"token_budget": {
"type": "integer",
"description": "Max tokens for response. Server truncates content to fit budget. Use memory(action='get') for full content of specific IDs. With 1M context models, budgets up to 100K are practical.",
@ -96,6 +106,10 @@ struct SearchArgs {
detail_level: Option<String>,
#[serde(alias = "context_topics")]
context_topics: Option<Vec<String>>,
#[serde(alias = "exclude_types")]
exclude_types: Option<Vec<String>>,
#[serde(alias = "include_types")]
include_types: Option<Vec<String>>,
#[serde(alias = "token_budget")]
token_budget: Option<i32>,
#[serde(alias = "retrieval_mode")]
@ -174,7 +188,14 @@ pub async fn execute(
let overfetch_limit = (limit * overfetch_multiplier).min(100); // Cap at 100 to avoid excessive DB load
let results = storage
.hybrid_search(&args.query, overfetch_limit, keyword_weight, semantic_weight)
.hybrid_search_filtered(
&args.query,
overfetch_limit,
keyword_weight,
semantic_weight,
args.include_types.as_deref(),
args.exclude_types.as_deref(),
)
.map_err(|e| e.to_string())?;
// Filter by min_retention and min_similarity first (cheap filters)