fix(core): preserve OMP reasoning and add host protocol contracts

.github/workflows/ci.yml

@@ -3,8 +3,7 @@ name: CI
 on:
   push:
     branches: [main]
-  pull_request:
-    branches: [main]
+  pull_request: {}
   workflow_dispatch:
   # The tag release calls these same gates for its exact commit.
   workflow_call:
@@ -30,7 +29,7 @@ jobs:
       # binary against the mock API and otherwise skips silently.
       - name: Install pi
         run: |
-          npm install -g @earendil-works/pi-coding-agent@latest
+          npm install -g @earendil-works/pi-coding-agent@1.1.0
           echo "PI_BIN=$(npm prefix -g)/bin/pi" >> "$GITHUB_ENV"
       - name: Verify pi starts
         run: '"$PI_BIN" --version'
@@ -69,14 +68,16 @@ jobs:
       - run: npm ci
       - name: Install Oh My Pi
         run: |
-          npm install -g @oh-my-pi/pi-coding-agent@latest
+          npm install -g @oh-my-pi/pi-coding-agent@18.11.1
           echo "OMP_BIN=$(npm prefix -g)/bin/omp" >> "$GITHUB_ENV"
       - name: Verify omp starts
         run: '"$OMP_BIN" --version'
       - name: Run OMP compatibility suite
         env:
           OMP_COMPAT_REQUIRED: "1"
-        run: node tests/test-omp-compat.mjs
+        run: |
+          node tests/test-omp-compat.mjs
+          COMMANDCODE_CONTRACT_HOST=omp npm run test:host-protocols
 
   # ──────────────────────────────────────────────────────────
   # Code-level vulnerability scanning (SAST)

.github/workflows/latest-hosts.yml

@@ -0,0 +1,36 @@
+name: Latest host compatibility
+
+on:
+  schedule:
+    - cron: "43 7 * * 1"
+  workflow_dispatch:
+
+permissions:
+  contents: read
+
+jobs:
+  latest:
+    runs-on: ubuntu-latest
+    timeout-minutes: 20
+    steps:
+      - uses: actions/checkout@v4
+      - uses: actions/setup-node@v4
+        with:
+          node-version: 22
+          cache: npm
+      - uses: oven-sh/setup-bun@v2.2.0
+        with:
+          bun-version: 1.4.0
+      - run: npm ci
+      - name: Install latest hosts (separate from pinned PR gates)
+        run: |
+          npm install -g @earendil-works/pi-coding-agent@latest @oh-my-pi/pi-coding-agent@latest
+          echo "PI_BIN=$(npm prefix -g)/bin/pi" >> "$GITHUB_ENV"
+          echo "OMP_BIN=$(npm prefix -g)/bin/omp" >> "$GITHUB_ENV"
+      - name: Record versions
+        run: |
+          { "$PI_BIN" --version; "$OMP_BIN" --version; node --version; bun --version; } | tee -a "$GITHUB_STEP_SUMMARY"
+      - run: npm test
+        env:
+          PI_LOCAL_REQUIRED: "1"
+          OMP_COMPAT_REQUIRED: "1"

CHANGELOG.md

@@ -2,6 +2,9 @@
 
 ## Unreleased
 
+- Preserve Chat Completions reasoning history on OMP with an explicit replay compatibility flag, verified by real Pi/OMP follow-up requests rather than only visible thinking deltas.
+- Add shared mock protocol contracts for reasoning, system prompts, parallel tools and image results across Chat Completions, Responses and Anthropic Messages, including packed-package checks. Pin supported hosts in PR CI and check latest hosts weekly. Allow explicit multi-model live profiles without enabling paid CI calls.
+
 ## 0.7.8 - 2026-10-15
 
 - Redact sensitive text in direct stream errors and normalized context-overflow messages before returning them to the host. This closes two error-reporting paths that did not use the shared redactor; it is defensive hardening, not a claim of a confirmed credential leak ([#155](https://github.com/patlux/pi-commandcode-provider/pull/155)).

CONTRIBUTING.md

@@ -67,6 +67,14 @@ npm install -g @oh-my-pi/pi-coding-agent
 OMP_BIN="$(npm prefix -g)/bin/omp" node tests/test-omp-compat.mjs
 ```
 
+### Protocol and host contracts
+
+`npm run test:host-protocols` runs both available hosts against loopback mocks. It verifies Chat Completions, Responses and Anthropic Messages, including system instructions, replayed reasoning/signatures, parallel tool-call/result pairing and image results on the follow-up wire. It uses temporary homes and a deterministic tool with no external side effects. Set `COMMANDCODE_CONTRACT_HOST=pi` or `omp` to select one host. Required-host environment variables make missing hosts fail rather than skip.
+
+PR gates pin Pi **1.1.0** and OMP **18.11.1**. The weekly `Latest host compatibility` workflow tests current releases separately; it does not silently move the supported baseline. Update pins only after both the complete suite and packed-package smoke pass. CI also runs for stacked PRs, not only PRs targeting main.
+
+Live model testing remains explicit and may incur costs. Set `COMMANDCODE_E2E_PROVIDER_MODELS` to a comma-separated list to test more than one model sequentially, for example `deepseek/deepseek-v4.1-flash,meta/muse-spark-1.3,claude-sonnet-4-6`. The selected models must be available to the account. Equivalent `_MODELS` variables work for each profile; do not combine `_MODEL` and `_MODELS`. No live credentials or paid calls are used by CI.
+
 Before opening a PR, run:
 
 ```sh

index.ts

@@ -111,7 +111,11 @@ function commandCodeHeaders(): Record<string, string> | undefined {
  */
 type CommandCodeProviderConfig = ProviderConfig & { usage?: UsageProvider }
 
-type CommandCodeModelCompat = NonNullable<ProviderConfig["models"]>[number]["compat"]
+// OMP requires an explicit opt-in to replay Chat Completions reasoning.
+// Pi preserves the streamed signature itself and ignores this extra flag.
+type CommandCodeModelCompat = NonNullable<ProviderConfig["models"]>[number]["compat"] & {
+  replayReasoningContent?: boolean
+}
 
 /**
  * Compatibility flags per wire. Chat Completions mirrors what the Provider API
@@ -140,6 +144,7 @@ function compatForModel(model: CommandCodeModel): CommandCodeModelCompat {
     supportsStore: false,
     supportsDeveloperRole: false,
     supportsReasoningEffort: MODEL_EFFORTS[model.id] !== undefined,
+    replayReasoningContent: model.reasoning,
     maxTokensField: "max_tokens",
   }
 }

package.json

@@ -29,7 +29,7 @@
     "LICENSE"
   ],
   "scripts": {
-    "test": "npm run typecheck && node --test tests/test-release.mjs tests/test-pi-child-cleanup.mjs && tsx tests/test-package-manifest.ts && tsx tests/test-api-key.ts && tsx tests/test-pure-functions.ts && tsx tests/test-models.ts && tsx tests/test-model-metadata-check.ts && tsx tests/test-runtime.ts && tsx tests/test-pricing.ts && tsx tests/test-pricing-check.ts && tsx tests/test-cost.ts && tsx tests/test-oauth.ts && tsx tests/test-abort.ts && tsx tests/test-overflow.ts && tsx tests/test-stream.ts && tsx tests/test-quota.ts && tsx tests/test-quota-command.ts && tsx tests/test-usage.ts && tsx tests/test-retry.ts && tsx tests/test-transport.ts && node tests/test-pi-isolated.mjs && node tests/test-pi-authenticated.mjs && node tests/test-pi-local.mjs && node tests/test-omp-compat.mjs",
+    "test": "npm run typecheck && node --test tests/test-release.mjs tests/test-pi-child-cleanup.mjs && tsx tests/test-package-manifest.ts && tsx tests/test-api-key.ts && tsx tests/test-pure-functions.ts && tsx tests/test-models.ts && tsx tests/test-model-metadata-check.ts && tsx tests/test-runtime.ts && tsx tests/test-pricing.ts && tsx tests/test-pricing-check.ts && tsx tests/test-cost.ts && tsx tests/test-oauth.ts && tsx tests/test-abort.ts && tsx tests/test-overflow.ts && tsx tests/test-stream.ts && tsx tests/test-quota.ts && tsx tests/test-quota-command.ts && tsx tests/test-usage.ts && tsx tests/test-retry.ts && tsx tests/test-transport.ts && node tests/test-pi-isolated.mjs && node tests/test-pi-authenticated.mjs && node tests/test-pi-local.mjs && node tests/test-omp-compat.mjs && node tests/test-host-protocols.mjs",
     "typecheck": "tsc --noEmit",
     "format:check": "prettier --check '**/*.{ts,mjs,json,md}'",
     "format": "prettier --write '**/*.{ts,mjs,json,md}'",
@@ -54,6 +54,7 @@
     "test:transport": "tsx tests/test-transport.ts",
     "test:pi-isolated": "node tests/test-pi-isolated.mjs",
     "test:pi-authenticated": "node tests/test-pi-authenticated.mjs",
+    "test:host-protocols": "node tests/test-host-protocols.mjs",
     "test:pi-local": "node tests/test-pi-local.mjs",
     "test:smoke": "node tests/test-smoke.mjs",
     "test:release": "node --test tests/test-release.mjs",

scripts/live-e2e-profile.mjs

@@ -38,16 +38,25 @@ async function credentialFor(profile) {
   throw new Error(`Set ${prefix}_API_KEY_FILE (recommended) or ${prefix}_API_KEY`)
 }
 
-function runProfile(profile, apiKey) {
+function modelsFor(profile) {
   const modelVariable =
     profile === "go"
       ? "COMMANDCODE_E2E_GO_MODEL"
       : profile === "goat"
         ? "COMMANDCODE_E2E_GOAT_MODEL"
         : "COMMANDCODE_E2E_PROVIDER_MODEL"
-  const model =
+  const multiple = process.env[`${modelVariable}S`]
+  if (multiple && process.env[modelVariable])
+    throw new Error(`${modelVariable} and ${modelVariable}S are mutually exclusive`)
+  const models = multiple?.split(",").map((model) => model.trim()) ?? [
     process.env[modelVariable] ??
-    (profile === "goat" ? "xai/grok-4.6" : "deepseek/deepseek-v4-flash")
+      (profile === "goat" ? "xai/grok-4.6" : "deepseek/deepseek-v4-flash"),
+  ]
+  if (models.some((model) => !model)) throw new Error(`Empty model in ${modelVariable}S`)
+  return [...new Set(models)]
+}
+
+function runProfile(profile, apiKey, model) {
   const env = {
     ...process.env,
     COMMAND_CODE_API_KEY: apiKey,
@@ -79,7 +88,9 @@ function runProfile(profile, apiKey) {
 
 try {
   for (const profile of profiles) {
-    await runProfile(profile, await credentialFor(profile))
+    const models = modelsFor(profile)
+    const credential = await credentialFor(profile)
+    for (const model of models) await runProfile(profile, credential, model)
   }
 } catch (error) {
   console.error(error instanceof Error ? error.message : String(error))

tests/fixtures/protocol-tools-extension.ts

@@ -0,0 +1,43 @@
+/** Deterministic test tools. No filesystem or network access. */
+interface FixtureTool {
+  name: string
+  label: string
+  description: string
+  parameters: unknown
+  execute: (
+    id: string,
+    args: { image?: string | null },
+  ) => Promise<{
+    content: ({ type: "text"; text: string } | { type: "image"; data: string; mimeType: string })[]
+    details: undefined
+  }>
+}
+interface FixtureApi {
+  registerTool(tool: FixtureTool): void
+}
+
+export default function (pi: FixtureApi) {
+  pi.registerTool({
+    name: "contract_read",
+    label: "Contract read",
+    description: "Return a synthetic image and text for protocol contract tests",
+    parameters: {
+      type: "object",
+      properties: { image: { type: ["string", "null"] } },
+      required: ["image"],
+    },
+    async execute(_id, args) {
+      return {
+        content: [
+          { type: "text", text: `contract-result-${args.image}` },
+          {
+            type: "image",
+            data: "iVBORw0KGgoAAAANSUhEUgAAAAEAAAABCAQAAAC1HAwCAAAAC0lEQVR42mP8/x8AAwMCAO+jRZkAAAAASUVORK5CYII=",
+            mimeType: "image/png",
+          },
+        ],
+        details: undefined,
+      }
+    },
+  })
+}