You signed in with another tab or window. Reload to refresh your session.You signed out in another tab or window. Reload to refresh your session.You switched accounts on another tab or window. Reload to refresh your session.Dismiss alert
{{ message }}
Repository navigation
Commit a279303
Browse filesBrowse the repository at this point in the historyBrowse files
- Final code candidate: `908be9cabb2b552c56cbd86537fbccb86ea5e0b2`
6
+
- Final code candidate: `071f4dc782e70bb4a0f63397902a4285d0156903`
7
7
- Mode: full Council review plus final independent remediation pass
8
8
- Local verdict: **PASS**
9
-
- Remote status: final commits still need to be pushed and fresh CI/bot review must finish
9
+
- Remote gate: the final candidate must be pushed and fresh CI/bot review must finish
10
10
11
11
## Decision
12
12
@@ -20,7 +20,7 @@ PDF accounting is intentionally approximate. The user selected a parser-free pol
20
20
PDF allowance = max(32,768, decoded inline payload bytes)
21
21
```
22
22
23
-
Remote URLs and provider file IDs receive the fixed allowance. Exact page expansion and tokenization remain provider-authoritative. The Council did not require exact PDF parsing because adding a parser would create a new document-processing/resource boundary without making provider token accounting exact.
23
+
Remote references and provider file IDs receive the fixed allowance when the part is identifiable as a PDF. An untyped provider file ID cannot be classified locally and receives the generic 16,384-token file allowance instead. Exact page expansion and tokenization remain provider-authoritative. The Council did not require exact PDF parsing because adding a parser would create a new document-processing/resource boundary without making provider token accounting exact.
24
24
25
25
## Original Council gate
26
26
@@ -44,7 +44,11 @@ Post-Council review found and repaired:
44
44
- small-model output floors above the model's own reservation;
45
45
- safety-margin loss near the context boundary;
46
46
- non-JSON provider-option loss during reasoning reconciliation;
47
-
- raw PDF page-marker trust.
47
+
- raw PDF page-marker trust;
48
+
- provider-normalized Mistral/Devstral bridge messages missing from the estimate;
49
+
- generated system prompts counted even when workflows omit them or OAuth routes them through instructions;
50
+
- identical system and instruction values being deduplicated despite occupying two wire fields; and
51
+
- a static bridge assertion that could match the unclamped property-access form.
48
52
49
53
Each repair was centralized in the shared output-budget module or the existing pure media projection, then exercised through both request boundaries.
50
54
@@ -61,40 +65,45 @@ Batching the same parser would not remove its decompression/traversal boundary.
61
65
62
66
## Final independent remediation review
63
67
64
-
Two independent live Council seats re-reviewed exact head `908be9cabb`after the parser removal.
68
+
Two independent live Council seats re-reviewed the request-shape correction at `c78e1a61b6`, then received the final `c78e1a61b6..071f4dc782` delta after fresh bot findings. The final delta makes the estimator more conservative and does not change PDF behavior.
65
69
66
70
### Feynman seat — PASS
67
71
68
-
- Verified exact final head.
72
+
- Verified exact final code head`071f4dc782`.
69
73
- Confirmed `pdf-lib`, async parsing, page regex/policy, and transitive lock entries are absent.
70
74
- Confirmed lazy estimation is not evaluated when `maxOutputTokens` is omitted.
71
-
- Confirmed both request boundaries clamp after headers, tools, instructions, and media projection are finalized.
- Confirmed the parser experiment remains cleanly reverted and no PDF dependency returned.
85
+
- Confirmed synchronous lazy estimation, linear Mistral projection, and parity across both callers.
86
+
- Confirmed identical wire fields are counted separately while OAuth/workflow omissions are not double-counted.
87
+
- Confirmed the fixed-width bridge lookbehind runs correctly under Bun.
88
+
- Confirmed focused tests, typecheck, and diff checks pass.
80
89
81
90
### Degraded seat
82
91
83
92
The original chairman seat was rejected twice by the service safety filter because its retained conversation context included the earlier parser stress case. It produced no contrary code finding on the final head. The thread limit prevented replacing that retained seat with a new fourth thread.
84
93
85
-
Consensus therefore rests on two independent live PASS votes, the primary review, the complete changed-file inspection, and a sealed zero-finding Codex Security remediation scan. The degraded seat is disclosed rather than silently counted as agreement.
94
+
Consensus therefore rests on two independent live PASS votes on exact final code head `071f4dc782`, the primary review, the complete changed-file inspection, and sealed zero-finding Codex Security scans. The degraded seat is disclosed rather than silently counted as agreement.
86
95
87
96
## Final verification
88
97
89
-
- Focused provider, AI-SDK stream, native request, and upstream bridge suites: **413 passed, 11 skipped, 1 existing todo, 0 failed**.
98
+
- Focused provider, AI-SDK stream, native request, and upstream bridge suites: **416 passed, 11 skipped, 1 existing todo, 0 failed**.
90
99
- Repository typecheck: **13/13 successful**.
91
100
- Strict changed-file marker validation: passed.
92
101
- Required-marker inventory: **35/35**.
93
102
- Frozen lockfile install: **1,295 installs across 1,390 packages, no changes**.
94
-
- Targeted oxlint: **212 warnings, 0 errors**.
103
+
- Targeted oxlint on the final supplemental files: **161 warnings, 0 errors**; warnings are existing repository debt.
95
104
- Prettier: all changed files pass except `provider/provider.ts` and `provider/transform.ts`; both fail identically on `origin/main`.
96
105
-`git diff --check origin/main...HEAD`: passed.
97
-
- Final Codex Security remediation scan: complete coverage, **0 findings**.
106
+
- Final request-shape Codex Security scans `1dcbce9e-587e-4495-94ff-6d3291e5e1d6` and `bfa1cc4a-7d87-463d-8fc9-822e1624f8b1`: complete coverage, **0 findings**.
- Findings remaining on the final candidate: **0**
@@ -16,7 +16,7 @@ The final candidate removes local PDF parsing entirely. PDF media now uses a del
16
16
max(32,768 tokens, decoded inline payload bytes)
17
17
```
18
18
19
-
Remote URLs and provider file IDs receive the fixed 32,768-token allowance. The estimate is monotonic in locally observable payload size, ignores untrusted page metadata, and does not decompress or traverse PDF structure. The configured provider remains authoritative for exact tokenization and for unusually compact or dense documents.
19
+
Remote references and provider file IDs receive the fixed 32,768-token allowance when the part is identifiable as a PDF. An untyped provider file ID cannot be classified locally and receives the generic 16,384-token file allowance instead. The estimate is monotonic in locally observable payload size, ignores untrusted page metadata, and does not decompress or traverse PDF structure. The configured provider remains authoritative for exact tokenization and for unusually compact or dense documents.
20
20
21
21
This is the selected product boundary, not an attempt at exact PDF accounting. A compact or dense PDF can still be underestimated locally and rejected by the provider. That residual is an acknowledged reliability limitation; it is not a local parser, authorization, confidentiality, integrity, or shared-service vulnerability.
22
22
@@ -31,8 +31,10 @@ The review used immutable ranges so every material repair was independently reco
The final remediation scan was sealed once as scan `80591880-0a17-454c-b312-92c32a38f5ff`. Its authoritative result contains six reviewed surfaces, no deferred work, no open question, and zero findings.
37
+
The parser-free remediation scan was sealed once as scan `80591880-0a17-454c-b312-92c32a38f5ff`. The final request-shape scans were sealed once each as `1dcbce9e-587e-4495-94ff-6d3291e5e1d6` and `bfa1cc4a-7d87-463d-8fc9-822e1624f8b1`. Their authoritative results contain no deferred work, no open question, and zero findings.
36
38
37
39
## Findings discovered and resolved
38
40
@@ -60,7 +62,7 @@ Both production request paths use the same sequence:
- Prettier: all changed files pass except `provider/provider.ts` and `provider/transform.ts`; both fail identically at `origin/main`, so no unrelated whole-file rewrite was introduced.
80
82
-`git diff --check origin/main...HEAD`: passed.
81
-
- Final Codex Security remediation scan: complete, **0 findings**.
83
+
- Final request-shape Codex Security scans: complete coverage, **0 findings**.
82
84
83
85
## Operational caveats
84
86
85
87
- TAC advisory was attempted once earlier in the PR workflow and was unavailable; it was not retried.
86
88
- No live provider request or full interactive UI replay was performed.
87
89
- Provider-side rejection remains possible for compact or unusually dense PDFs because the local estimate is intentionally crude.
88
-
- The remote PR still needs the final commits pushed, current review threads reconciled, and fresh CI/bot results before a merge recommendation.
90
+
- The remote PR must receive the final commits, reconcile every current review thread, and pass fresh CI/bot review before merge.
0 commit comments