@functional-systems/srcx 0.1.0-dev.31 → 0.1.0-dev.32

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 CHANGED
@@ -453,6 +453,12 @@ The subsequent [exact-Collection lookup replay](https://github.com/lambdadb/srcx
453
453
  removes 22 requests per command in that project, with identical source results
454
454
  and lower observed read/search latency. It does not measure complete-agent cost.
455
455
 
456
+ In the subsequent [agent choice comparison](https://github.com/lambdadb/srcx/blob/develop/eval/VERSION-AGENT-RESULTS.md),
457
+ all three agents offered srcx read its installed skill but used local tools only
458
+ to investigate the prepared CLI/SDK versions. No srcx retrieval benefit was
459
+ established; multiple repositories and historical versions alone did not lead to
460
+ tool adoption when their source trees were already available locally.
461
+
456
462
  The internal [lexical chunking pilot](https://github.com/lambdadb/srcx/blob/develop/eval/README.md) compares syntax-aware and
457
463
  window chunks on a fixed public corpus and 16 pinned investigation queries.
458
464
  `npm run eval:prepare` builds and validates the comparison offline. Live evaluation
package/VALIDATION.md CHANGED
@@ -1,5 +1,25 @@
1
1
  # Validation record
2
2
 
3
+ ## Agent choice across repositories and versions
4
+
5
+ The [six-session comparison](eval/VERSION-AGENT-RESULTS.md) offers the installed
6
+ srcx skill alongside identical local CLI/SDK source trees for three fresh
7
+ maintenance questions. All three srcx-condition agents read the skill but make
8
+ zero srcx calls. Both conditions meet 14 frozen facts and partially meet one;
9
+ no material answer errors were found under assistant source review. All 82 linked
10
+ citation ranges were opened and checked. Tests cited by agents were not executed.
11
+
12
+ Mean total input is 11.0% lower and time 8.1% lower with srcx offered, but uncached
13
+ input is 6.1% higher. With no retrieval calls, these are session differences, not
14
+ evidence of srcx search efficiency or cost savings. Stop tuning this prepared-local
15
+ workflow; any team pilot needs an actual recurring source-access need.
16
+
17
+ All six sessions finish within budget, all 18 source checkouts stay clean at their
18
+ pins, and all 31 frozen inputs match. Six before/after resolution checks confirm
19
+ unchanged commit/Tag/Snapshot identities. No imports, task searches, embeddings or
20
+ remote writes occur. The protocol, rubric and runtime hashes were committed before
21
+ execution; raw sessions and source evidence remain in the local evaluation archive.
22
+
3
23
  ## Direct Collection lookup
4
24
 
5
25
  Exact generated Collection selectors now fetch only the selected metadata and
package/package.json CHANGED
@@ -1,6 +1,6 @@
1
1
  {
2
2
  "name": "@functional-systems/srcx",
3
- "version": "0.1.0-dev.31",
3
+ "version": "0.1.0-dev.32",
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": "642169ee7abdc8fc71508418f43a52ca863dce5f"
69
+ "gitHead": "88abc665f4f436c045d79870eeb7cf0eac8ac814"
70
70
  }