@functional-systems/srcx 0.1.0-dev.28 → 0.1.0-dev.29
This diff represents the content of publicly available package versions that have been released to one of the supported registries. The information contained in this diff is provided for informational purposes only and reflects changes between package versions as they appear in their respective public registries.
- package/README.md +5 -0
- package/VALIDATION.md +17 -0
- package/package.json +2 -2
package/README.md
CHANGED
|
@@ -439,6 +439,11 @@ Agents can use the following search guidance:
|
|
|
439
439
|
|
|
440
440
|
## Evaluate retrieval quality
|
|
441
441
|
|
|
442
|
+
The [fresh agent navigation comparison](https://github.com/lambdadb/srcx/blob/develop/eval/FRESH-AGENT-NAVIGATION-RESULTS.md)
|
|
443
|
+
measures complete answers, model input and wall time with local tools available in
|
|
444
|
+
both conditions. It does not establish a clear efficiency advantage on a small
|
|
445
|
+
local repository; retrieval quality and agent-workflow benefits remain separate.
|
|
446
|
+
|
|
442
447
|
The internal [lexical chunking pilot](https://github.com/lambdadb/srcx/blob/develop/eval/README.md) compares syntax-aware and
|
|
443
448
|
window chunks on a fixed public corpus and 16 pinned investigation queries.
|
|
444
449
|
`npm run eval:prepare` builds and validates the comparison offline. Live evaluation
|
package/VALIDATION.md
CHANGED
|
@@ -1,5 +1,22 @@
|
|
|
1
1
|
# Validation record
|
|
2
2
|
|
|
3
|
+
## Fresh adaptive agent navigation
|
|
4
|
+
|
|
5
|
+
[Eight fresh sessions](eval/FRESH-AGENT-NAVIGATION-RESULTS.md) compare local tools
|
|
6
|
+
against local tools plus srcx on four new Requests behavior questions. Both arms
|
|
7
|
+
support all frozen facts under assistant source review. Mean total input is 3.2%
|
|
8
|
+
lower with srcx, while mean wall time is 6.3% longer; one observation per condition
|
|
9
|
+
does not establish a general efficiency advantage. Three lexical searches and one
|
|
10
|
+
semantic search were followed by local reads; no srcx implementation read occurred.
|
|
11
|
+
|
|
12
|
+
All 18 previews/handles match pinned source and version metadata, eight checkouts
|
|
13
|
+
remain clean, and 31 frozen inputs match their pre-run hashes. Setup/audit resolution
|
|
14
|
+
confirms the same immutable Snapshot. Argument and exhausted-budget guard checks
|
|
15
|
+
passed before execution. Two recoverable local command errors remain in the data;
|
|
16
|
+
all sessions and service calls completed. No reindexing, document embeddings,
|
|
17
|
+
ranking changes or agent reruns were performed. Raw task workspaces were archived
|
|
18
|
+
with all 1,310 file hashes verified; historical evaluations remain unchanged.
|
|
19
|
+
|
|
3
20
|
## Optional overload implementation reads
|
|
4
21
|
|
|
5
22
|
[Offline navigation results](eval/IMPLEMENTATION-READ-RESULTS.md) cover six fresh
|
package/package.json
CHANGED
|
@@ -1,6 +1,6 @@
|
|
|
1
1
|
{
|
|
2
2
|
"name": "@functional-systems/srcx",
|
|
3
|
-
"version": "0.1.0-dev.
|
|
3
|
+
"version": "0.1.0-dev.29",
|
|
4
4
|
"private": false,
|
|
5
5
|
"description": "Version-aware code search on LambdaDB",
|
|
6
6
|
"type": "module",
|
|
@@ -66,5 +66,5 @@
|
|
|
66
66
|
"url": "https://github.com/lambdadb/srcx/issues"
|
|
67
67
|
},
|
|
68
68
|
"license": "Apache-2.0",
|
|
69
|
-
"gitHead": "
|
|
69
|
+
"gitHead": "b8985da513109e33668492688988772b172f5080"
|
|
70
70
|
}
|