feat(ai-reviews): bump ai-workflows to v3.1.5 + switch auto-review to DeepSeek V3.2

[sfreudenthaler]

Two changes to get automatic PR reviews working again and keep them legible:

1. Switch automatic reviewer GPT-5.5 → DeepSeek V3.2

openai.gpt-5.5 is failing on Bedrock Mantle in both us-east-1 and us-east-2 as of 2026-06-15 (invalid_prompt: 404 Not Found: Engine not found — AWS-side; the alias and its -2026-04-23 snapshot both 404 while still listed in /v1/models). Raised with our AWS TAM.

deepseek.v3.2 is healthy on bedrock-runtime (Converse) and is still a non-Claude reviewer, so it preserves the model-diversity rationale (Claude writes the code, a different family reviews it). deepseek.* routes to the existing bedrock-generic (Converse) executor — no IAM or ai-workflows change needed (BedrockInvokeReviewModels already allows bedrock:Converse on foundation-model/deepseek.*).

• Renamed gpt-automatic-review → ai-automatic-review (model-agnostic); dropped reasoning_effort (mantle-only).
• Interactive @claude and the backend reviewer stay on Anthropic Claude.
• Validated e2e on steve-quarterly-planning Cannot add roles to "Who can use" on workflows #108 (routed to bedrock-generic, posted a real review catching all planted bugs).

2. Bump pin v3.1.4 → v3.1.5

ai-workflows#40: surfaces the actual failure reason on response.failed (e.g. server_error / Engine not found) in the sticky + job log instead of a generic "failed before producing output", and retries once on transient failures. This is exactly what made the gpt-5.5 outage diagnosable.

Switch back to a mantle model by setting model_id to an openai.* id once AWS restores gpt-5.5.

Closes: #36181

[claude]

Claude finished @sfreudenthaler's task in 45s —— View job

---

Rollback Safety Analysis

• Gather context and understand the PR
• Read rollback-unsafe categories reference
• Get full PR diff
• Analyze diff against unsafe categories
• Apply appropriate label

---

Result: ✅ Safe To Rollback

The diff touches only two GitHub Actions workflow files:

• .github/workflows/ai_claude-backend-reviewer.yml — bumps @v3.1.4 → @v3.1.5
• .github/workflows/ai_claude-orchestrator.yml — bumps @v3.1.4 → @v3.1.5, renames the gpt-automatic-review job to ai-automatic-review, switches model_id from openai.gpt-5.5 to deepseek.v3.2, removes reasoning_effort: medium, and updates inline comments

None of the ROLLBACK_UNSAFE_CATEGORIES apply:

Category	Match?
C-1 Structural Data Model Change	❌ No DB/storage changes
C-2 ES Mapping Change	❌ No Elasticsearch changes
C-3 Content JSON Model Version	❌ No Java model changes
C-4 DROP TABLE / DROP COLUMN	❌ No schema changes
H-1 through H-8 (data/API/OSGi)	❌ No Java/REST/VTL changes
M-1 through M-4 (column types, API, OSGi)	❌ No applicable changes

These are pure CI/CD workflow changes. Rolling back to the previous release leaves the application binary, database schema, Elasticsearch index, and all API contracts completely unchanged. Label AI: Safe To Rollback has been applied.

[github-actions]

❌ Codex Review — model service error — openai.gpt-5.5

⚠️ Codex could not produce a review — the model service returned an error: invalid_prompt: Task submission failed with status 404 Not Found: Engine not found

This is typically an AWS / bedrock-mantle service-side issue (e.g. server_error), not a problem with this PR. It usually clears on its own — re-run the job to retry.

---

_{Run: #27586538700 · tokens: in: ? · out: ? (reasoning: ?) · total: ?}

[github-actions]

🤖 Bedrock Review — deepseek.v3.2

[🟡 Medium] .github/workflows/ai_claude-backend-reviewer.yml:88 — Updates workflow to use claude-orchestrator.yml@v3.1.5. Ensure this new version exists and is compatible.

[🟡 Medium] .github/workflows/ai_claude-orchestrator.yml:85 — Updates workflow to use claude-orchestrator.yml@v3.1.5. Ensure this new version exists and is compatible.

[🟡 Medium] .github/workflows/ai_claude-orchestrator.yml:148 — Updates workflow to use claude-orchestrator.yml@v3.1.5. Ensure this new version exists and is compatible.

[🟡 Medium] .github/workflows/ai_claude-orchestrator.yml:152 — Switches model from openai.gpt-5.5 to deepseek.v3.2. Verify the new model ID is correct and functional on the AWS Bedrock runtime path.

[🟡 Medium] .github/workflows/ai_claude-orchestrator.yml:153 — Removes reasoning_effort: medium. Confirm this parameter is not required for the new deepseek.v3.2 model or the generic Converse API route.

---

_{Run: #27587619190 · tokens: in: 1394 · out: 280 · total: 1674}

[wezell]

let's see it work

[sfreudenthaler]

let's see it work

got confirmation from AWS TAM that gpt 5.5 was busted on their end (they just don't have a public place to find out). that's why it was working then stopped.

v3.1.5 of ai-workflows fixes the problem that it was swallowing the error as well.