Skip to content

Commit ea10cd7

Browse files
norandomclaude
andcommitted
docs(research): R107 gate P4 of the model seam; gate assessed as proceed; design revised
Co-Authored-By: Claude <noreply@anthropic.com>
1 parent fea6344 commit ea10cd7

18 files changed

Lines changed: 752 additions & 12 deletions

‎.kiro/specs/model-seam/design.md‎

Lines changed: 11 additions & 4 deletions
Original file line numberDiff line numberDiff line change
@@ -1,8 +1,8 @@
11
# Design: the model seam
22

33
Generated 2026-11-12 (Europe/Berlin) from the approved [requirements](requirements.md).
4-
**Provisional.** Design approval is pending the [prototype gate](prototype-gate.md); no production tasks are
5-
generated before it is assessed as `proceed`. Discovery log: [research.md](research.md).
4+
Revised 2026-11-12 after the [prototype gate](prototype-gate.md) was assessed as `proceed` (R105, R106, R107).
5+
Awaiting the owner's approval; tasks are generated after it. Discovery log: [research.md](research.md).
66

77
## Overview
88

@@ -476,8 +476,15 @@ Each step leaves the repository working. Step 4 is the only one that changes the
476476

477477
## Prototype gate
478478

479-
See [prototype-gate.md](prototype-gate.md): P1 library comparison, P2 build on five targets, P3 waiting,
480-
P4 process boundary and receipts. Status: not run.
479+
See [prototype-gate.md](prototype-gate.md). Assessed 2026-11-12 as `proceed`: P1 passed (one library on
480+
all six first cases), P2 passed (five targets, either TLS option), P3 passed on a stub, P4 passed for the
481+
process boundary and earlier receipts. Two items are carried into implementation, each as a task with
482+
its check: one real refusal waited out with its announced estimate (3.1 to 3.4 live), and a new receipt
483+
that states only what was sent (8.4), shown once the hand-written request has left the core.
484+
485+
Found in the gate and taken into this design: the seam returns a reply's text as the model wrote it, and
486+
the core removes one code fence around it before parsing; the Anthropic style's address gets its version
487+
segment when the endpoint file leaves it out.
481488

482489
## Contradictions and disposition
483490

‎.kiro/specs/model-seam/prototype-gate.md‎

Lines changed: 10 additions & 4 deletions
Original file line numberDiff line numberDiff line change
@@ -1,6 +1,7 @@
11
# Prototype gate: model-seam
22

3-
Written 2026-11-12 with the provisional design. Status: **not run** except where stated. The design stays
3+
Written 2026-11-12 with the provisional design. Status: **proceed** (assessment at the end), run on the
4+
same day. The design stays
45
provisional and design approval stays pending until this gate is assessed as `proceed`
56
(`.kiro/steering/prototype-first.md`). Results go to dated entries under `docs/research/` and are linked
67
here.
@@ -56,8 +57,6 @@ headers are reachable, licence, and the date of the latest release.
5657

5758
## Assessment
5859

59-
Not assessed as a whole.
60-
6160
- P1, 2026-11-12: **passes** for `genai` 0.7.4 on all six first cases; a refusal's status, body and
6261
headers are reachable. The alternative `llm` 1.3.8 completes none of the five OpenAI-style cases.
6362
[R105](../../../docs/research/2026-11-12/12-model-seam-gate-p1.md).
@@ -67,4 +66,11 @@ Not assessed as a whole.
6766
- P3, 2026-11-12: **passes on the stub** (five scripted cases). Its live condition is **not met**: no
6867
refusal could be provoked on the local endpoint with 38 calls at once. Kept open; to be closed when a
6968
long run meets a rate limit. R106.
70-
- P4: not run.
69+
- P4, 2026-11-12: **passes for the process boundary and earlier receipts**: `regenerate run` completes on
70+
all six first cases through a second executable; 390 tests pass on the prototype branch
71+
(`model-seam-gate-p4`, `37effda`). **Not met:** the new receipt also carries settings of the unused
72+
hand-written request. [R107](../../../docs/research/2026-11-12/14-model-seam-gate-p4.md).
73+
74+
**Gate as a whole, 2026-11-12: `proceed`.** Carried into implementation as named items, each with a task
75+
and a check: (1) one real refusal waited out with its announced estimate; (2) a new receipt states only
76+
what was sent, shown after the hand-written request has left the core.

‎.kiro/specs/model-seam/spec.json‎

Lines changed: 6 additions & 3 deletions
Original file line numberDiff line numberDiff line change
@@ -1,7 +1,7 @@
11
{
22
"feature_name": "model-seam",
33
"created_at": "2026-11-12T17:45:00+01:00",
4-
"updated_at": "2026-11-12T18:26:00+01:00",
4+
"updated_at": "2026-11-12T20:10:00+01:00",
55
"language": "en",
66
"phase": "design-generated",
77
"approvals": {
@@ -20,7 +20,7 @@
2020
},
2121
"ready_for_implementation": false,
2222
"prototype_gate": "prototype-gate.md",
23-
"prototype_status": "not-run",
23+
"prototype_status": "proceed",
2424
"related": [
2525
"constraint-guided-compression",
2626
"deconvolution-loop",
@@ -30,7 +30,10 @@
3030
"evidence": [
3131
"docs/research/2026-11-12/06-repair-gate-result.md",
3232
"docs/research/2026-11-12/09-product-path-two-models.md",
33-
"docs/research/2026-11-12/10-model-seam-library-prototype.md"
33+
"docs/research/2026-11-12/10-model-seam-library-prototype.md",
34+
"docs/research/2026-11-12/12-model-seam-gate-p1.md",
35+
"docs/research/2026-11-12/13-model-seam-gate-p2-p3.md",
36+
"docs/research/2026-11-12/14-model-seam-gate-p4.md"
3437
],
3538
"notes": "2026-11-12: owner statements: language packs are published once and not regenerated; adopters must be able to run GLM or another model behind the tool; core and driver model agnostic, fence because inference is not deterministic; use an existing model library; preferred solution is horizontal compatibility with an estimated wait time; include the internal endpoint on localhost. Discovery brief written and the owner said 'go ahead' to requirements. Requirements generated 2026-11-12 and approved by the owner the same day (\"approved\"), including the four proposed points: the estimate is of a call's wait, not of a whole run; ten minutes total wait per call unless set otherwise; at most two repairs; the check command makes at most three calls. Provisional design, research log and prototype gate (P1 to P4) written the same day; design approval pending the gate; two open decisions for the owner in the design's contradiction table (rows 1 and 5)."
3639
}
Lines changed: 5 additions & 0 deletions
Original file line numberDiff line numberDiff line change
@@ -0,0 +1,5 @@
1+
def total_price(items):
2+
total = 0
3+
for item in items:
4+
total = total + item["price"] * item["count"]
5+
return total
Lines changed: 5 additions & 0 deletions
Original file line numberDiff line numberDiff line change
@@ -0,0 +1,5 @@
1+
def total_price(items):
2+
total = 0
3+
for item in items:
4+
total = total + item["price"] * item["count"]
5+
return total
Lines changed: 5 additions & 0 deletions
Original file line numberDiff line numberDiff line change
@@ -0,0 +1,5 @@
1+
def total_price(items):
2+
total = 0
3+
for item in items:
4+
total = total + item["price"] * item["count"]
5+
return total
Lines changed: 2 additions & 0 deletions
Original file line numberDiff line numberDiff line change
@@ -0,0 +1,2 @@
1+
def total_price(items):
2+
return sum(item["price"] * item["count"] for item in items)
Lines changed: 5 additions & 0 deletions
Original file line numberDiff line numberDiff line change
@@ -0,0 +1,5 @@
1+
def total_price(items):
2+
total = 0
3+
for item in items:
4+
total = total + item["price"] * item["count"]
5+
return total
Lines changed: 2 additions & 0 deletions
Original file line numberDiff line numberDiff line change
@@ -0,0 +1,2 @@
1+
def total_price(items):
2+
return sum(item["price"] * item["count"] for item in items)
Lines changed: 94 additions & 0 deletions
Original file line numberDiff line numberDiff line change
@@ -0,0 +1,94 @@
1+
{
2+
"facts": [
3+
{
4+
"apparatus": {
5+
"regenerator.rs": "4ebd7741a1f273fad3d8e3a35be24361a186af24be76451ed758a9a7503e4a53"
6+
},
7+
"candidate_root": "D:\\BuildArtifacts\\lesscode\\model-seam\\cand-p4-1",
8+
"endpoint": {
9+
"host": "api.mistral.ai",
10+
"model": "zai-glm-5-3",
11+
"model_sha256": "6bc336046b01392684b855540e81e7b90481f212bccf27d8dc939f43bb4a1144",
12+
"style": "openai"
13+
},
14+
"feedback_kind": null,
15+
"frozen_scope": {
16+
"src/mod.py": "6b450b456b325b6ba0e81355c773977c831cd1f7a40ab4abc31c6935e850ca6b"
17+
},
18+
"key": "p4-1-inputs",
19+
"language": "python",
20+
"preregistered_params": {
21+
"effort": "high",
22+
"max_tokens": 64000,
23+
"rollout_id": 1,
24+
"seed": null,
25+
"temperature": null
26+
},
27+
"prompts": {
28+
"system.txt": "7a3719269af6d8643a2a0e5256d8e5170bec59820f24d24912c9ce94664fa6ee",
29+
"user.txt": "e39d1f85b2c2f2e7c32f520845f8a10ec86c5c868b77355c8c3d203f39ea4bcf"
30+
},
31+
"request": {
32+
"D:\\BuildArtifacts\\lesscode\\model-seam\\request.json": "1d06538f2efb22cfda808cbddc089981bf5dd272377112ef55ea21bbe84db0dd"
33+
},
34+
"surface": "ab6b54e2978d53a10c1db5983b4712ee25ed937b2da8b7e3ba43f7a184da2fdb",
35+
"time_utc": "2026-11-12T19:08:16.078842+00:00"
36+
}
37+
],
38+
"notes": [
39+
{
40+
"key": "p4-1-turn",
41+
"time_utc": "2026-11-12T19:08:17.688361+00:00",
42+
"value": {
43+
"effort": "high",
44+
"elapsed_s": 1.61,
45+
"max_tokens": 64000,
46+
"mode": "openai-json-http",
47+
"model_returned": "zai-glm-5-3",
48+
"model_sent": "zai-glm-5-3",
49+
"request_sha256": "fbc68e4298eed92e969a51e8dbfd4871526711e8872fac5c7b25dae8710073e3",
50+
"response_sha256": "b4d7be9b3531510e3d0519232222dd56ecf481bf3f00fe52a525a9dc1ae59922",
51+
"response_sha256_saved": "b4d7be9b3531510e3d0519232222dd56ecf481bf3f00fe52a525a9dc1ae59922",
52+
"seed": null,
53+
"seed_note": "endpoint exposes no seed; determinism not claimed",
54+
"stop_reason": "stop",
55+
"system_sha256": "7a3719269af6d8643a2a0e5256d8e5170bec59820f24d24912c9ce94664fa6ee",
56+
"temperature": 0.0,
57+
"usage": {
58+
"completion_tokens": 142,
59+
"prompt_tokens": 645,
60+
"seconds": 1.5,
61+
"sent": {
62+
"host": "api.mistral.ai",
63+
"max_output_tokens": 64000,
64+
"model": "zai-glm-5-3",
65+
"reasoning_level": "high",
66+
"temperature": null
67+
},
68+
"waits": []
69+
},
70+
"user_sha256": "e39d1f85b2c2f2e7c32f520845f8a10ec86c5c868b77355c8c3d203f39ea4bcf"
71+
}
72+
},
73+
{
74+
"time_utc": "2026-11-12T19:08:17.690992+00:00",
75+
"key": "p4-1-candidate",
76+
"value": {
77+
"root": "D:\\BuildArtifacts\\lesscode\\model-seam\\cand-p4-1",
78+
"files": [
79+
{
80+
"path": "src/mod.py",
81+
"bytes": 132,
82+
"sha256": "fc7644cec2212966bacdd03bc27a11daed87c9cdee198397077f003df71fc82c",
83+
"crlf": false
84+
}
85+
],
86+
"result_sha256": "ac242aa6f244a1bd3acfc35eb346bd768f0186d9dd181e865aa37031aa9857f8",
87+
"structure_counts": {
88+
"pure": 1,
89+
"shell": 0
90+
}
91+
}
92+
}
93+
]
94+
}

0 commit comments

Comments
 (0)