AI search engine and MCP server for bots and agents. One Rust binary indexes code, docs, logs, JSONL/CSV and PDFs for BM25 full-text, vector and hybrid search, code search without grep, agent memory and RAG. Elasticsearch-compatible API; Corpus Hub serves hundreds of reference memory corpora.
Rust agent-memoryai-agentsbm25claude-codeclicode-searchcursordatabaseelasticsearchfull-text-searchhybrid-searchjsonllocal-firstmcpmcp-servermodel-context-protocolragrustsearch-enginevector-search
Merge pull request #327 from probelabs/feat/definition-aware-code-search
xerj-orga9013f7ac3ae
diff --cc engine/crates/xerj-api/src/auth.rsindex b00abbd58,046c42b1a..1a8b1fd1e--- a/engine/crates/xerj-api/src/auth.rs+++ b/engine/crates/xerj-api/src/auth.rs@@@ -460,17 -420,10 +460,18 @@@ fn unauthorized_response(data_dir: &strrequest_id: None,},status: 401,+ xerj_hint: None,};- (StatusCode::UNAUTHORIZED, Json(body)).into_response()+ (+ StatusCode::UNAUTHORIZED,+ [(+ axum::http::header::WWW_AUTHENTICATE,+ WWW_AUTHENTICATE_CHALLENGE,+ )],+ Json(body),+ )+ .into_response()}// ─────────────────────────────────────────────────────────────────────────────diff --cc engine/crates/xerj-autoindex/src/lib.rsindex b3800c64a,3e3dec046..a0ead54d7--- a/engine/crates/xerj-autoindex/src/lib.rs+++ b/engine/crates/xerj-autoindex/src/lib.rs@@@ -4857,15 -4694,56 +4857,62 @@@ pub fn run_index_report(cfg: IndexCfg));}println!(- "\nnext: `xerj autoindex map --url {}` for the data map; search via GET /{}-*/_search",- cfg.url, cfg.prefix+ "{}",+ next_hint(+ &cfg.url,+ &cfg.prefix,+ cfg.api_key.as_deref(),+ std::env::var("XERJ_API_KEY").ok().as_deref(),+ cfg.api_key_file.as_deref(),+ ));+ // This line is where an agent learns how to search, so it is where the+ // code-aware form belongs. A blind usability run showed the cost of+ // omitting it: the agent followed exactly this guidance, got whole+ // files back, and reinvented client-side slicing rather than+ // discovering `_passage` — which it never saw mentioned anywhere.+ // Only printed when source code was actually indexed; for a PDF or log+ // corpus it would be noise.+ let indexed_code = plan.files.values().any(|fa| fa.family == "code");+ if indexed_code {+ println!(+ " code was indexed — for a function/class instead of a whole file:\n\+ \x20 curl -s $URL/{}-*/_search -H 'Content-Type: application/json' \\\n\+ \x20 -d '{{\"query\":{{\"bool\":{{\"should\":[{{\"multi_match\":{{\"query\":\"<symbol or phrase>\",\+ \"fields\":[\"body\",\"defs\"],\"type\":\"most_fields\"}}}},{{\"match_phrase\":{{\"defs\":{{\"query\":\"<symbol or phrase>\",\"boost\":4}}}}}}]}}}},\+ \"_source\":[\"ax_path\",\"title\"],\"fields\":[\"_passage\"]}}'\n\+ \x20 (the `match_phrase` clause ranks the file that DEFINES a symbol above files that merely call it; `_passage` returns \+ the enclosing block, not the file)",+ cfg.prefix+ );+ }+ // A mixed corpus needs one more thing said out loud. Code volume+ // swamps prose on shared vocabulary: in a blind run, a question about+ // documentation ("exit code 3") matched WordPress PHP on `exit` and+ // `code` and returned nothing but source, costing a wasted round trip.+ // The agent worked out the fix itself afterwards — scope the query —+ // which is exactly the sort of thing that should not have to be+ // rediscovered.+ let indexed_prose = plan+ .files+ .values()+ .any(|fa| fa.family.starts_with("txt") || fa.family == "html");+ if indexed_code && indexed_prose {+ println!(+ " this corpus mixes code and prose — code volume can swamp prose on\n\+ \x20 shared words. Scope the side you want:\n\+ \x20 …,\"query\":{{\"bool\":{{\"must\":[{{\"match\":{{\"body\":\"<text>\"}}}}],\+ \"filter\":[{{\"term\":{{\"ax_format\":\"code\"}}}}]}}}}\n\+ \x20 (values for `ax_format` in this run: {})",+ {+ let mut fams: Vec<&str> =+ plan.files.values().map(|fa| fa.family.as_str()).collect();+ fams.sort_unstable();+ fams.dedup();+ fams.join(", ")+ }+ );+ }}// Exit 3 means "completed, some input was unusable". Backend rejections// are not consulted and must not be: a rejected item aborts the run withdiff --cc engine/crates/xerj-server/src/main.rsindex cf026cfd3,558bfb9a1..12ac6c373--- a/engine/crates/xerj-server/src/main.rs+++ b/engine/crates/xerj-server/src/main.rs@@@ -247,11 -254,19 +265,23 @@@ fn help_text() -> Stringxerj autoindex map print the discovered data map\n\xerj brain <folder> one command: index a folder into a running, browsable\n\second brain in your browser (see xerj brain --help)\n\+ xerj mcp [opts] Model Context Protocol stdio server: exposes 10 tools\n\+ (search, semantic, vector, hybrid, memory, second-brain)\n\+ to any MCP client. Proxies to a node you already started\n\+ — set XERJ_URL or --url (see xerj mcp --help)\n\\n\+ SEARCHING A CODEBASE:\n\+ A search hit returns the whole indexed document, which for source code is the\n\+ whole file. To get only the matching function or class, ask for the passage:\n\+ \n\+ curl -s $URL/<index>/_search -H 'Content-Type: application/json' \\\n\+ -d '{{\"query\":{{\"bool\":{{\"should\":[{{\"multi_match\":{{\"query\":\"<symbol or phrase>\",\n\+ \"fields\":[\"body\",\"defs\"],\"type\":\"most_fields\"}}}},{{\"match_phrase\":{{\"defs\":{{\"query\":\"<symbol or phrase>\",\"boost\":4}}}}}}]}}}},\n\+ \"_source\":[\"ax_path\",\"title\"],\"fields\":[\"_passage\"]}}'\n\+ \n\+ the `match_phrase` clause ranks the file that DEFINES a symbol above files that merely call it;\n\+ `_passage` returns the enclosing block. Without them you get whole files.\n\+ \n\ENVIRONMENT:\n\XERJ_LOG Log level filter (default: info)\n\XERJ_CONFIG Config file path\n\