deploy(bedrock): grant instance-role Bedrock invoke + repoint LLM_MODEL_ID / eval model ids

Deployment-readiness for the Bedrock migration (PR #62):
- instance-role.ts: least-privilege bedrock:InvokeModel[WithResponseStream] on the
  us.anthropic.claude-opus-4-8 inference-profile ARN + the foundation-model ARN in
  each routed region (us-east-1/2, us-west-2). The model runs in the server process
  on the box, so the EC2 instance role is the principal. Simulator-verified (allowed
  for opus-4-8, implicitDeny for other models) and synth-verified. Passed the
  mandatory GPT-4.1 IAM cross-review (no blockers, least-privilege confirmed).
- config-store.ts: IaC SSM LLM_MODEL_ID anthropic:claude-opus-4-8 ->
  bedrock_converse:us.anthropic.claude-opus-4-8. This SSM value overrides
  seed_store.sh's default via pick precedence, so the seed-script fix alone was
  insufficient — both sources now point at the supported Bedrock id.
- infra/README.md + evals/reviewer/config.toml: repoint stale anthropic:/google_genai:
  ids to the Bedrock id (config.toml's model_id was an active, now-broken value).

AWS_REGION is already wired via user-data.sh (IMDS -> boot.env), so no change needed there.

evals/reviewer/config.toml

@@ -9,8 +9,9 @@ langsmith_project = "open-swe-evals"
 # Leave blank to use LANGGRAPH_URL or local dev.
 langgraph_url = ""
 assistant_id = "reviewer"
-# models: openai:gpt-5.5, anthropic:claude-opus-4-8, google_genai:gemini-3.5-flash
-model_id = "google_genai:gemini-3.5-flash"
+# models (post Bedrock/Fireworks migration): bedrock_converse:us.anthropic.claude-opus-4-8,
+# or any fireworks:* id in agent/dashboard/options.py SUPPORTED_MODELS.
+model_id = "bedrock_converse:us.anthropic.claude-opus-4-8"
 reasoning_effort = "medium"
 
 # score_mode:

infra/README.md

@@ -156,7 +156,7 @@ Three buckets:
    | `DASHBOARD_BASE_URL` | `https://openswe-dev.seahaven.com` *(confirm host)* | `https://openswe.seahaven.com` *(confirm host)* |
    | `DASHBOARD_API_BASE_URL` | same as base | same as base |
    | `DASHBOARD_ALLOWED_ORIGINS` | same as base | same as base |
-   | `LLM_MODEL_ID` | `anthropic:claude-opus-4-8` *(confirm)* | `anthropic:claude-opus-4-8` *(confirm)* |
+   | `LLM_MODEL_ID` | `bedrock_converse:us.anthropic.claude-opus-4-8` | `bedrock_converse:us.anthropic.claude-opus-4-8` |
 
 3. **Out-of-band SSM config — NOT created by CDK.** Operationally-variable or
    env-specific-unknown values listed in `OUT_OF_BAND_SSM` and populated by

infra/lib/constructs/config-store.ts

@@ -142,8 +142,12 @@ export function iacManagedSsm(env: EnvName): Record<string, string> {
     DASHBOARD_BASE_URL: host,
     DASHBOARD_API_BASE_URL: host,
     DASHBOARD_ALLOWED_ORIGINS: host,
-    // Primary builder model (project memory team_settings: anthropic:claude-opus-4-8).
-    LLM_MODEL_ID: "anthropic:claude-opus-4-8",
+    // Primary builder model. seed_store.sh's `pick` precedence is
+    // OPENSWE_AGENT_MODEL > SEED_AGENT_MODEL > LLM_MODEL_ID > script default, so this
+    // SSM value overrides the seed-script default — it MUST be a supported id. Post
+    // Bedrock/Fireworks migration the only Bedrock-Claude id is the inference profile;
+    // `anthropic:claude-opus-4-8` was removed from SUPPORTED_MODELS.
+    LLM_MODEL_ID: "bedrock_converse:us.anthropic.claude-opus-4-8",
   };
   // Dev e2e smoke: seed the owner's user_mapping so an @openswe comment from the
   // triggering GitHub login resolves (an unmapped commenter is silently skipped).

infra/lib/constructs/instance-role.ts

@@ -117,6 +117,29 @@ export class InstanceRole extends Construct {
     // moves these to a customer CMK, add a scoped `kms:Decrypt` on that key ARN
     // ONLY (not `*`).
 
+    // Invoke the Bedrock Claude model. DEFAULT_MODEL_ID is
+    // `bedrock_converse:us.anthropic.claude-opus-4-8`, and the model runs in the
+    // LangGraph server PROCESS on this box (not in the sandbox), so the EC2
+    // instance role is the calling principal. The `us.` cross-region inference
+    // profile fans out to us-east-1 / us-east-2 / us-west-2, and Bedrock authorizes
+    // InvokeModel against BOTH the inference-profile ARN AND the underlying
+    // foundation-model ARN in each routed region — all four resources are required
+    // or the call AccessDenies. Scoped to opus-4-8 ONLY (least-privilege): adding a
+    // new Bedrock model to SUPPORTED_MODELS means extending this resource list.
+    // IAM change — flag for T4 (GPT-4.1 IAM cross-review) / T5 (/sh-security-review).
+    this.role.addToPolicy(
+      new iam.PolicyStatement({
+        sid: "InvokeBedrockClaude",
+        actions: ["bedrock:InvokeModel", "bedrock:InvokeModelWithResponseStream"],
+        resources: [
+          `arn:aws:bedrock:${REGION}:${ACCOUNT}:inference-profile/us.anthropic.claude-opus-4-8`,
+          "arn:aws:bedrock:us-east-1::foundation-model/anthropic.claude-opus-4-8",
+          "arn:aws:bedrock:us-east-2::foundation-model/anthropic.claude-opus-4-8",
+          "arn:aws:bedrock:us-west-2::foundation-model/anthropic.claude-opus-4-8",
+        ],
+      }),
+    );
+
     // Ship application logs to CloudWatch Logs under /open-swe/<env>/*.
     this.role.addToPolicy(
       new iam.PolicyStatement({