Fase 8 · Bloco B — suíte de correção (property + golden + SSE-correctness) (#3808)

Integrated into release/v3.8.25 — Fase 8 Bloco B (property + golden + SSE-correctness).
2026-07-26 09:52:11 +03:00 · 2026-06-14 18:02:51 -03:00
parent cf5898205a
commit 931afe3482
34 changed files with 1054 additions and 12 deletions
--- a/.github/workflows/ci.yml
+++ b/.github/workflows/ci.yml
@@ -348,7 +348,7 @@ jobs:
          cache: npm
      - run: npm ci
      - run: npm run check:node-runtime
-      - run: node --max-old-space-size=4096 --import tsx --test --test-force-exit --test-concurrency=4 --test-shard=${{ matrix.shard }}/8 tests/unit/*.test.ts "tests/unit/{api,auth,authz,build,cli,cli-helper,compression,cors,dashboard,db,db-adapters,docs,gamification,guardrails,lib,mcp,runtime,security,services,settings,shared,ui}/**/*.test.ts"
+      - run: node --max-old-space-size=4096 --import tsx --test --test-force-exit --test-concurrency=4 --test-shard=${{ matrix.shard }}/8 tests/unit/*.test.ts "tests/unit/{api,auth,authz,build,cli,cli-helper,compression,correctness,cors,dashboard,db,db-adapters,docs,gamification,guardrails,lib,mcp,runtime,security,services,settings,shared,ui}/**/*.test.ts"

  test-vitest:
    name: Vitest (MCP / autoCombo / UI components)
@@ -397,7 +397,7 @@ jobs:
      - run: npm ci
      - run: npm run check:node-runtime
      - run: npm run build
-      - run: node --import tsx --test --test-force-exit --test-concurrency=4 --test-shard=${{ matrix.shard }}/2 tests/unit/*.test.ts "tests/unit/{api,auth,authz,build,cli,cli-helper,compression,cors,dashboard,db,db-adapters,docs,gamification,guardrails,lib,mcp,runtime,security,services,settings,shared,ui}/**/*.test.ts"
+      - run: node --import tsx --test --test-force-exit --test-concurrency=4 --test-shard=${{ matrix.shard }}/2 tests/unit/*.test.ts "tests/unit/{api,auth,authz,build,cli,cli-helper,compression,correctness,cors,dashboard,db,db-adapters,docs,gamification,guardrails,lib,mcp,runtime,security,services,settings,shared,ui}/**/*.test.ts"

  node-26-compat:
    name: Node 26 Compatibility (${{ matrix.shard }}/2)
@@ -421,7 +421,7 @@ jobs:
      - run: npm ci
      - run: npm run check:node-runtime
      - run: npm run build
-      - run: node --import tsx --test --test-force-exit --test-concurrency=4 --test-shard=${{ matrix.shard }}/2 tests/unit/*.test.ts "tests/unit/{api,auth,authz,build,cli,cli-helper,compression,cors,dashboard,db,db-adapters,docs,gamification,guardrails,lib,mcp,runtime,security,services,settings,shared,ui}/**/*.test.ts"
+      - run: node --import tsx --test --test-force-exit --test-concurrency=4 --test-shard=${{ matrix.shard }}/2 tests/unit/*.test.ts "tests/unit/{api,auth,authz,build,cli,cli-helper,compression,correctness,cors,dashboard,db,db-adapters,docs,gamification,guardrails,lib,mcp,runtime,security,services,settings,shared,ui}/**/*.test.ts"

  test-coverage-shard:
    name: Coverage Shard (${{ matrix.shard }}/8)
@@ -460,7 +460,7 @@ jobs:
            --exclude=tests/** \
            --exclude=**/*.test.* \
            node --max-old-space-size=4096 --import tsx --test --test-force-exit --test-concurrency=4 \
-              --test-shard=${{ matrix.shard }}/8 tests/unit/*.test.ts "tests/unit/{api,auth,authz,build,cli,cli-helper,compression,cors,dashboard,db,db-adapters,docs,gamification,guardrails,lib,mcp,runtime,security,services,settings,shared,ui}/**/*.test.ts"
+              --test-shard=${{ matrix.shard }}/8 tests/unit/*.test.ts "tests/unit/{api,auth,authz,build,cli,cli-helper,compression,correctness,cors,dashboard,db,db-adapters,docs,gamification,guardrails,lib,mcp,runtime,security,services,settings,shared,ui}/**/*.test.ts"
      - name: Upload raw shard coverage
        if: always()
        uses: actions/upload-artifact@v7
--- a/.github/workflows/nightly-property.yml
+++ b/.github/workflows/nightly-property.yml
@@ -0,0 +1,33 @@
+name: Nightly Property Discovery
+on:
+  schedule:
+    - cron: "0 6 * * *"
+  workflow_dispatch:
+permissions:
+  contents: read
+  issues: write
+jobs:
+  property-random-seed:
+    runs-on: ubuntu-latest
+    steps:
+      - uses: actions/checkout@v6
+      - uses: actions/setup-node@v6
+        with:
+          node-version: "24"
+          cache: npm
+      - run: npm ci
+      - name: fast-check random seed (high runs)
+        id: prop
+        run: FC_SEED=random FC_NUM_RUNS=2000 npm run test:property
+      - name: Open issue on failure
+        if: failure()
+        uses: actions/github-script@v7
+        with:
+          script: |
+            await github.rest.issues.create({
+              owner: context.repo.owner,
+              repo: context.repo.repo,
+              title: "Nightly property-test failure (Fase 8 B)",
+              body: "fast-check found a counterexample with a random seed. Check the run logs for the reproducible seed + minimal case, then add it as a fixture.\n\nRun: " + context.serverUrl + "/" + context.repo.owner + "/" + context.repo.repo + "/actions/runs/" + context.runId,
+              labels: ["quality-gate-finding"],
+            });
--- a/dependency-allowlist.json
+++ b/dependency-allowlist.json
@@ -45,6 +45,7 @@
    "eslint-config-next",
    "eslint-plugin-sonarjs",
    "express",
+    "fast-check",
    "fetch-socks",
    "fflate",
    "fumadocs-core",
--- a/package-lock.json
+++ b/package-lock.json
@@ -109,6 +109,7 @@
        "eslint": "^9.39.4",
        "eslint-config-next": "16.2.7",
        "eslint-plugin-sonarjs": "^4.0.3",
+        "fast-check": "^4.8.0",
        "glob": "^13.0.6",
        "husky": "^9.1.7",
        "jscpd": "^4.2.5",
@@ -13489,6 +13490,29 @@
        "node": ">=0.10.0"
      }
    },
+    "node_modules/fast-check": {
+      "version": "4.8.0",
+      "resolved": "https://registry.npmjs.org/fast-check/-/fast-check-4.8.0.tgz",
+      "integrity": "sha512-GOJ158CUMnN6cSahsv4+ExARvIDuzzinFjkp0E9WtiBa5zcVeLozVkWaE4IzFcc+Y48Wp1EDlUZsXRyAztQcSg==",
+      "dev": true,
+      "funding": [
+        {
+          "type": "individual",
+          "url": "https://github.com/sponsors/dubzzz"
+        },
+        {
+          "type": "opencollective",
+          "url": "https://opencollective.com/fast-check"
+        }
+      ],
+      "license": "MIT",
+      "dependencies": {
+        "pure-rand": "^8.0.0"
+      },
+      "engines": {
+        "node": ">=12.17.0"
+      }
+    },
    "node_modules/fast-copy": {
      "version": "4.0.2",
      "resolved": "https://registry.npmjs.org/fast-copy/-/fast-copy-4.0.2.tgz",
@@ -21758,6 +21782,23 @@
        "url": "https://github.com/sponsors/sindresorhus"
      }
    },
+    "node_modules/pure-rand": {
+      "version": "8.4.0",
+      "resolved": "https://registry.npmjs.org/pure-rand/-/pure-rand-8.4.0.tgz",
+      "integrity": "sha512-IoM8YF/jY0hiugFo/wOWqfmarlE6J0wc6fDK1PhftMk7MGhVZl88sZimmqBBFomLOCSmcCCpsfj7wXASCpvK9A==",
+      "dev": true,
+      "funding": [
+        {
+          "type": "individual",
+          "url": "https://github.com/sponsors/dubzzz"
+        },
+        {
+          "type": "opencollective",
+          "url": "https://opencollective.com/fast-check"
+        }
+      ],
+      "license": "MIT"
+    },
    "node_modules/pvtsutils": {
      "version": "1.3.6",
      "resolved": "https://registry.npmjs.org/pvtsutils/-/pvtsutils-1.3.6.tgz",
--- a/package.json
+++ b/package.json
@@ -84,15 +84,16 @@
    "electron:build:mac": "npm run build && cd electron && npm run build:mac",
    "electron:build:linux": "npm run build && cd electron && npm run build:linux",
    "electron:smoke:packaged": "node scripts/dev/smoke-electron-packaged.mjs",
-    "test": "cross-env DISABLE_SQLITE_AUTO_BACKUP=true node --max-old-space-size=8192 --import tsx --import ./open-sse/utils/setupPolyfill.ts --test --test-concurrency=20 tests/unit/*.test.ts \"tests/unit/{api,auth,authz,build,cli,cli-helper,compression,cors,dashboard,db,db-adapters,docs,gamification,guardrails,lib,mcp,runtime,security,services,settings,shared,ui}/**/*.test.ts\"",
-    "test:unit": "cross-env DISABLE_SQLITE_AUTO_BACKUP=true node --max-old-space-size=8192 --import tsx --import ./open-sse/utils/setupPolyfill.ts --test --test-force-exit --test-concurrency=20 tests/unit/*.test.ts \"tests/unit/{api,auth,authz,build,cli,cli-helper,compression,cors,dashboard,db,db-adapters,docs,gamification,guardrails,lib,mcp,runtime,security,services,settings,shared,ui}/**/*.test.ts\"",
-    "test:unit:fast": "cross-env DISABLE_SQLITE_AUTO_BACKUP=true node --max-old-space-size=8192 --import tsx --import ./open-sse/utils/setupPolyfill.ts --test --test-force-exit --test-isolation=none tests/unit/*.test.ts \"tests/unit/{api,auth,authz,build,cli,cli-helper,compression,cors,dashboard,db,db-adapters,docs,gamification,guardrails,lib,mcp,runtime,security,services,settings,shared,ui}/**/*.test.ts\"",
+    "test": "cross-env DISABLE_SQLITE_AUTO_BACKUP=true node --max-old-space-size=8192 --import tsx --import ./open-sse/utils/setupPolyfill.ts --test --test-concurrency=20 tests/unit/*.test.ts \"tests/unit/{api,auth,authz,build,cli,cli-helper,compression,correctness,cors,dashboard,db,db-adapters,docs,gamification,guardrails,lib,mcp,runtime,security,services,settings,shared,ui}/**/*.test.ts\"",
+    "test:unit": "cross-env DISABLE_SQLITE_AUTO_BACKUP=true node --max-old-space-size=8192 --import tsx --import ./open-sse/utils/setupPolyfill.ts --test --test-force-exit --test-concurrency=20 tests/unit/*.test.ts \"tests/unit/{api,auth,authz,build,cli,cli-helper,compression,correctness,cors,dashboard,db,db-adapters,docs,gamification,guardrails,lib,mcp,runtime,security,services,settings,shared,ui}/**/*.test.ts\"",
+    "test:unit:fast": "cross-env DISABLE_SQLITE_AUTO_BACKUP=true node --max-old-space-size=8192 --import tsx --import ./open-sse/utils/setupPolyfill.ts --test --test-force-exit --test-isolation=none tests/unit/*.test.ts \"tests/unit/{api,auth,authz,build,cli,cli-helper,compression,correctness,cors,dashboard,db,db-adapters,docs,gamification,guardrails,lib,mcp,runtime,security,services,settings,shared,ui}/**/*.test.ts\"",
    "test:unit:shard": "concurrently --kill-others-on-fail -n s1,s2 \"npm:test:unit:shard:1\" \"npm:test:unit:shard:2\"",
-    "test:unit:shard:1": "cross-env DISABLE_SQLITE_AUTO_BACKUP=true node --max-old-space-size=8192 --import tsx --import ./open-sse/utils/setupPolyfill.ts --test --test-force-exit --test-concurrency=10 --test-shard=1/2 tests/unit/*.test.ts \"tests/unit/{api,auth,authz,build,cli,cli-helper,compression,cors,dashboard,db,db-adapters,docs,gamification,guardrails,lib,mcp,runtime,security,services,settings,shared,ui}/**/*.test.ts\"",
-    "test:unit:shard:2": "cross-env DISABLE_SQLITE_AUTO_BACKUP=true node --max-old-space-size=8192 --import tsx --import ./open-sse/utils/setupPolyfill.ts --test --test-force-exit --test-concurrency=10 --test-shard=2/2 tests/unit/*.test.ts \"tests/unit/{api,auth,authz,build,cli,cli-helper,compression,cors,dashboard,db,db-adapters,docs,gamification,guardrails,lib,mcp,runtime,security,services,settings,shared,ui}/**/*.test.ts\"",
+    "test:unit:shard:1": "cross-env DISABLE_SQLITE_AUTO_BACKUP=true node --max-old-space-size=8192 --import tsx --import ./open-sse/utils/setupPolyfill.ts --test --test-force-exit --test-concurrency=10 --test-shard=1/2 tests/unit/*.test.ts \"tests/unit/{api,auth,authz,build,cli,cli-helper,compression,correctness,cors,dashboard,db,db-adapters,docs,gamification,guardrails,lib,mcp,runtime,security,services,settings,shared,ui}/**/*.test.ts\"",
+    "test:unit:shard:2": "cross-env DISABLE_SQLITE_AUTO_BACKUP=true node --max-old-space-size=8192 --import tsx --import ./open-sse/utils/setupPolyfill.ts --test --test-force-exit --test-concurrency=10 --test-shard=2/2 tests/unit/*.test.ts \"tests/unit/{api,auth,authz,build,cli,cli-helper,compression,correctness,cors,dashboard,db,db-adapters,docs,gamification,guardrails,lib,mcp,runtime,security,services,settings,shared,ui}/**/*.test.ts\"",
    "test:plan3": "cross-env DISABLE_SQLITE_AUTO_BACKUP=true node --import tsx --import ./open-sse/utils/setupPolyfill.ts --test tests/unit/plan3-p0.test.ts",
    "test:fixes": "cross-env DISABLE_SQLITE_AUTO_BACKUP=true node --import tsx --import ./open-sse/utils/setupPolyfill.ts --test tests/unit/fixes-p1.test.ts",
    "test:security": "cross-env DISABLE_SQLITE_AUTO_BACKUP=true node --import tsx --import ./open-sse/utils/setupPolyfill.ts --test tests/unit/security-fase01.test.ts",
+    "test:property": "cross-env DISABLE_SQLITE_AUTO_BACKUP=true node --import tsx --test --test-force-exit tests/unit/correctness/*.property.test.ts",
    "check:cycles": "node scripts/check/check-cycles.mjs",
    "check:route-validation:t06": "node scripts/check/check-route-validation.mjs",
    "check:any-budget:t11": "node scripts/check/check-t11-any-budget.mjs",
@@ -164,7 +165,7 @@
    "test:vitest:ui": "vitest run --config vitest.config.ts tests/unit/ui",
    "test:ecosystem": "node scripts/dev/run-ecosystem-tests.mjs",
    "test:system": "cross-env DISABLE_SQLITE_AUTO_BACKUP=true node --import tsx --import ./open-sse/utils/setupPolyfill.ts --test --test-force-exit --test-concurrency=1 tests/e2e/system-failover.test.ts",
-    "test:coverage": "cross-env DISABLE_SQLITE_AUTO_BACKUP=true c8 --output-dir coverage --exclude=tests/** --exclude=**/*.test.* --reporter=text-summary --reporter=html --reporter=json-summary --reporter=lcov --check-coverage --statements 60 --lines 60 --functions 60 --branches 60 node --max-old-space-size=8192 --import tsx --import ./open-sse/utils/setupPolyfill.ts --test --test-force-exit --test-concurrency=8 tests/unit/*.test.ts \"tests/unit/{api,auth,authz,build,cli,cli-helper,compression,cors,dashboard,db,db-adapters,docs,gamification,guardrails,lib,mcp,runtime,security,services,settings,shared,ui}/**/*.test.ts\"",
+    "test:coverage": "cross-env DISABLE_SQLITE_AUTO_BACKUP=true c8 --output-dir coverage --exclude=tests/** --exclude=**/*.test.* --reporter=text-summary --reporter=html --reporter=json-summary --reporter=lcov --check-coverage --statements 60 --lines 60 --functions 60 --branches 60 node --max-old-space-size=8192 --import tsx --import ./open-sse/utils/setupPolyfill.ts --test --test-force-exit --test-concurrency=8 tests/unit/*.test.ts \"tests/unit/{api,auth,authz,build,cli,cli-helper,compression,correctness,cors,dashboard,db,db-adapters,docs,gamification,guardrails,lib,mcp,runtime,security,services,settings,shared,ui}/**/*.test.ts\"",
    "test:coverage:legacy": "c8 --output-dir coverage --exclude=open-sse --check-coverage --lines 50 --functions 50 --branches 50 node --import tsx --test tests/unit/*.test.ts",
    "coverage:report": "c8 report --output-dir coverage --exclude=tests/** --exclude=**/*.test.* --reporter=text --reporter=text-summary --reporter=html --reporter=json-summary --reporter=lcov",
    "coverage:summary": "node scripts/check/test-report-summary.mjs --input coverage/coverage-summary.json --output coverage/coverage-report.md",
@@ -279,6 +280,7 @@
    "eslint": "^9.39.4",
    "eslint-config-next": "16.2.7",
    "eslint-plugin-sonarjs": "^4.0.3",
+    "fast-check": "^4.8.0",
    "glob": "^13.0.6",
    "husky": "^9.1.7",
    "jscpd": "^4.2.5",
--- a/scripts/check/check-test-discovery.mjs
+++ b/scripts/check/check-test-discovery.mjs
@@ -53,7 +53,7 @@ export const COLLECTORS = [
  // "vitest" e explodem no node runner). Subdir novo: adicione aqui E nos scripts
  // (o drift-check + o gate de órfãos forçam a manutenção em sincronia).
  {
-    glob: "tests/unit/{api,auth,authz,build,cli,cli-helper,compression,cors,dashboard,db,db-adapters,docs,gamification,guardrails,lib,mcp,runtime,security,services,settings,shared,ui}/**/*.test.ts",
+    glob: "tests/unit/{api,auth,authz,build,cli,cli-helper,compression,correctness,cors,dashboard,db,db-adapters,docs,gamification,guardrails,lib,mcp,runtime,security,services,settings,shared,ui}/**/*.test.ts",
    sources: ["package.json", ".github/workflows/ci.yml"],
  },
  // Node native runner — test:integration (top-level only; tests/integration/services/ NÃO roda)
@@ -79,7 +79,10 @@ export const COLLECTORS = [
  { glob: "tests/e2e/*.spec.ts", sources: ["package.json"] },
  // Runners custom — test:ecosystem / test:protocols:e2e (spawnam vitest com o arquivo)
  { glob: "tests/e2e/ecosystem.test.ts", sources: ["scripts/dev/run-ecosystem-tests.mjs"] },
-  { glob: "tests/e2e/protocol-clients.test.ts", sources: ["scripts/dev/run-protocol-clients-tests.mjs"] },
+  {
+    glob: "tests/e2e/protocol-clients.test.ts",
+    sources: ["scripts/dev/run-protocol-clients-tests.mjs"],
+  },
 ];

 const escapeRe = (s) => s.replace(/[.*+?^${}()|[\]\\]/g, "\\$&");
--- a/scripts/quality/property-seed.mjs
+++ b/scripts/quality/property-seed.mjs
@@ -0,0 +1,3 @@
+// Fixed seed for PR runs (deterministic). Nightly overrides via FC_SEED=random.
+export const PROPERTY_SEED = 42424242;
+export const PROPERTY_NUM_RUNS = 200;
--- a/tests/fixtures/translation/claude-to-openai.json
+++ b/tests/fixtures/translation/claude-to-openai.json
@@ -0,0 +1,23 @@
+[
+  {
+    "name": "claude-to-openai/basic-message",
+    "sourceFormat": "claude",
+    "targetFormat": "openai",
+    "input": {
+      "model": "gpt-4o",
+      "messages": [{ "role": "user", "content": [{ "type": "text", "text": "Hello" }] }],
+      "max_tokens": 512
+    }
+  },
+  {
+    "name": "claude-to-openai/with-system",
+    "sourceFormat": "claude",
+    "targetFormat": "openai",
+    "input": {
+      "model": "gpt-4o-mini",
+      "system": [{ "type": "text", "text": "Be helpful." }],
+      "messages": [{ "role": "user", "content": [{ "type": "text", "text": "What is 1+1?" }] }],
+      "max_tokens": 256
+    }
+  }
+]
--- a/tests/fixtures/translation/gemini-to-openai.json
+++ b/tests/fixtures/translation/gemini-to-openai.json
@@ -0,0 +1,24 @@
+[
+  {
+    "name": "gemini-to-openai/basic-chat",
+    "sourceFormat": "gemini",
+    "targetFormat": "openai",
+    "input": {
+      "model": "gpt-4o",
+      "contents": [{ "role": "user", "parts": [{ "text": "Hello from Gemini format!" }] }]
+    }
+  },
+  {
+    "name": "gemini-to-openai/multi-turn",
+    "sourceFormat": "gemini",
+    "targetFormat": "openai",
+    "input": {
+      "model": "gpt-4o-mini",
+      "contents": [
+        { "role": "user", "parts": [{ "text": "What is 2+2?" }] },
+        { "role": "model", "parts": [{ "text": "4" }] },
+        { "role": "user", "parts": [{ "text": "And 3+3?" }] }
+      ]
+    }
+  }
+]
--- a/tests/fixtures/translation/openai-to-claude.json
+++ b/tests/fixtures/translation/openai-to-claude.json
@@ -0,0 +1,37 @@
+[
+  {
+    "name": "openai-to-claude/basic-chat",
+    "sourceFormat": "openai",
+    "targetFormat": "claude",
+    "input": {
+      "model": "claude-opus-4-5",
+      "messages": [{ "role": "user", "content": "hi" }]
+    }
+  },
+  {
+    "name": "openai-to-claude/with-system",
+    "sourceFormat": "openai",
+    "targetFormat": "claude",
+    "input": {
+      "model": "claude-sonnet-4-6",
+      "messages": [
+        { "role": "system", "content": "You are a helpful assistant." },
+        { "role": "user", "content": "Hello!" }
+      ],
+      "max_tokens": 1024
+    }
+  },
+  {
+    "name": "openai-to-claude/multi-turn",
+    "sourceFormat": "openai",
+    "targetFormat": "claude",
+    "input": {
+      "model": "claude-haiku-4-5",
+      "messages": [
+        { "role": "user", "content": "What is 2+2?" },
+        { "role": "assistant", "content": "4" },
+        { "role": "user", "content": "And 3+3?" }
+      ]
+    }
+  }
+]
--- a/tests/fixtures/translation/openai-to-gemini.json
+++ b/tests/fixtures/translation/openai-to-gemini.json
@@ -0,0 +1,24 @@
+[
+  {
+    "name": "openai-to-gemini/basic-chat",
+    "sourceFormat": "openai",
+    "targetFormat": "gemini",
+    "input": {
+      "model": "gemini-2.5-flash",
+      "messages": [{ "role": "user", "content": "Hello Gemini!" }]
+    }
+  },
+  {
+    "name": "openai-to-gemini/with-system",
+    "sourceFormat": "openai",
+    "targetFormat": "gemini",
+    "input": {
+      "model": "gemini-2.5-pro",
+      "messages": [
+        { "role": "system", "content": "You are a helpful assistant." },
+        { "role": "user", "content": "Tell me a joke." }
+      ],
+      "temperature": 0.7
+    }
+  }
+]
--- a/tests/fixtures/translation/sse-chunk-sequences.json
+++ b/tests/fixtures/translation/sse-chunk-sequences.json
@@ -0,0 +1,26 @@
+[
+  {
+    "name": "openai-stream/two-deltas-then-done",
+    "chunks": [
+      "data: {\"choices\":[{\"delta\":{\"content\":\"He\"}}]}\n\n",
+      "data: {\"choices\":[{\"delta\":{\"content\":\"llo\"}}]}\n\n",
+      "data: [DONE]\n\n"
+    ],
+    "expectedText": "Hello"
+  },
+  {
+    "name": "openai-stream/single-delta-then-done",
+    "chunks": ["data: {\"choices\":[{\"delta\":{\"content\":\"World\"}}]}\n\n", "data: [DONE]\n\n"],
+    "expectedText": "World"
+  },
+  {
+    "name": "openai-stream/multiple-deltas",
+    "chunks": [
+      "data: {\"choices\":[{\"delta\":{\"content\":\"foo\"}}]}\n\n",
+      "data: {\"choices\":[{\"delta\":{\"content\":\" \"}}]}\n\n",
+      "data: {\"choices\":[{\"delta\":{\"content\":\"bar\"}}]}\n\n",
+      "data: [DONE]\n\n"
+    ],
+    "expectedText": "foo bar"
+  }
+]
--- a/tests/helpers/fakeUpstreamStream.ts
+++ b/tests/helpers/fakeUpstreamStream.ts
@@ -0,0 +1,28 @@
+/**
+ * Controllable fake upstream ReadableStream for integration tests.
+ * Allows pushing chunks, closing, erroring, and observing cancel.
+ */
+export function fakeUpstreamStream() {
+  let controllerRef: ReadableStreamDefaultController<Uint8Array> | null = null;
+  let cancelCb: ((reason?: unknown) => void) | null = null;
+  const enc = new TextEncoder();
+
+  const stream = new ReadableStream<Uint8Array>({
+    start(controller) {
+      controllerRef = controller;
+    },
+    cancel(reason) {
+      cancelCb?.(reason);
+    },
+  });
+
+  return {
+    stream,
+    push: (s: string) => controllerRef?.enqueue(enc.encode(s)),
+    close: () => controllerRef?.close(),
+    error: (e: unknown) => controllerRef?.error(e),
+    onCancel: (cb: (reason?: unknown) => void) => {
+      cancelCb = cb;
+    },
+  };
+}
--- a/tests/helpers/goldenSnapshot.ts
+++ b/tests/helpers/goldenSnapshot.ts
@@ -0,0 +1,56 @@
+import fs from "node:fs";
+import path from "node:path";
+import { fileURLToPath } from "node:url";
+
+const DEFAULT_DIR = path.resolve(path.dirname(fileURLToPath(import.meta.url)), "../snapshots");
+
+/** Stable JSON serialization: object keys sorted recursively. */
+function stable(v: unknown): string {
+  return JSON.stringify(
+    v,
+    (_k, val) =>
+      val && typeof val === "object" && !Array.isArray(val)
+        ? Object.fromEntries(
+            Object.keys(val as Record<string, unknown>)
+              .sort()
+              .map((k) => [k, (val as Record<string, unknown>)[k]])
+          )
+        : val,
+    2
+  );
+}
+
+class GoldenMismatchError extends Error {
+  constructor(name: string, expected: string, actual: string) {
+    super(
+      `golden mismatch for "${name}". Run with UPDATE_GOLDEN=1 to regenerate.\n--- expected\n${expected}\n--- actual\n${actual}`
+    );
+    this.name = "GoldenMismatchError";
+  }
+}
+
+/**
+ * Compare `value` with a stable JSON snapshot at `tests/snapshots/<name>.json`.
+ * - First run (file missing): writes the file and passes.
+ * - UPDATE_GOLDEN=1: rewrites the file and passes.
+ * - Mismatch: throws GoldenMismatchError with a diff-friendly message.
+ *
+ * @param name - Slash-separated path under tests/snapshots/ (e.g. "translation/openai-to-claude/basic-chat")
+ * @param value - The value to serialize and compare.
+ * @param dir - Optional override for snapshot root directory (used in tests to isolate to tmpdir).
+ */
+export function goldenSnapshot(name: string, value: unknown, dir = DEFAULT_DIR): void {
+  const file = path.join(dir, `${name}.json`);
+  const serialized = stable(value);
+
+  if (process.env.UPDATE_GOLDEN === "1" || !fs.existsSync(file)) {
+    fs.mkdirSync(path.dirname(file), { recursive: true });
+    fs.writeFileSync(file, serialized + "\n");
+    return;
+  }
+
+  const expected = fs.readFileSync(file, "utf8").trimEnd();
+  if (serialized !== expected) {
+    throw new GoldenMismatchError(name, expected, serialized);
+  }
+}
--- a/tests/helpers/propertyConfig.ts
+++ b/tests/helpers/propertyConfig.ts
@@ -0,0 +1,13 @@
+import fc from "fast-check";
+import { PROPERTY_SEED, PROPERTY_NUM_RUNS } from "../../scripts/quality/property-seed.mjs";
+
+const envSeed = process.env.FC_SEED;
+export const seed = envSeed === "random" ? undefined : envSeed ? Number(envSeed) : PROPERTY_SEED;
+export const numRuns = process.env.FC_NUM_RUNS
+  ? Number(process.env.FC_NUM_RUNS)
+  : PROPERTY_NUM_RUNS;
+
+/** Apply the shared deterministic config; call once at top of each property test file. */
+export function configureProperties(): void {
+  fc.configureGlobal({ seed, numRuns });
+}
--- a/tests/helpers/translationFixtures.ts
+++ b/tests/helpers/translationFixtures.ts
@@ -0,0 +1,31 @@
+import fs from "node:fs";
+import path from "node:path";
+import { fileURLToPath } from "node:url";
+
+const DIR = path.resolve(path.dirname(fileURLToPath(import.meta.url)), "../fixtures/translation");
+
+export type TranslationCase = {
+  name: string;
+  sourceFormat: string;
+  targetFormat: string;
+  input: Record<string, unknown>;
+  expected?: unknown;
+};
+
+export type SseSequence = {
+  name: string;
+  chunks: string[];
+  expectedText: string;
+};
+
+export function loadTranslationFixtures(): TranslationCase[] {
+  return ["openai-to-claude", "claude-to-openai", "openai-to-gemini", "gemini-to-openai"].flatMap(
+    (f) => JSON.parse(fs.readFileSync(path.join(DIR, `${f}.json`), "utf8")) as TranslationCase[]
+  );
+}
+
+export function loadSseSequences(): SseSequence[] {
+  return JSON.parse(
+    fs.readFileSync(path.join(DIR, "sse-chunk-sequences.json"), "utf8")
+  ) as SseSequence[];
+}
--- a/tests/integration/sse-correctness.test.ts
+++ b/tests/integration/sse-correctness.test.ts
@@ -0,0 +1,120 @@
+/**
+ * SSE-correctness integration tests (Task 11, Fase 8 B).
+ *
+ * Drives the real createSSEStream pipeline through a controllable fake upstream.
+ * Run with: node --import tsx/esm --test --test-concurrency=1 tests/integration/sse-correctness.test.ts
+ *
+ * Notes on observed createSSEStream behavior (calibrated invariants):
+ * - The TransformStream processes SSE events and emits translated chunks to the client.
+ * - [DONE] is consumed by the pipeline: it closes the upstream readable and the
+ *   TransformStream flushes + terminates, but does NOT re-emit "data: [DONE]" to the output.
+ *   The output stream closes naturally (reader.read() returns {done:true}).
+ * - Upstream errors propagate as TransformStream errors (reader.read() rejects).
+ * - Cancel propagates via ReadableStream cancel callback (pipeThrough wires it).
+ */
+
+import test from "node:test";
+import assert from "node:assert/strict";
+import { fakeUpstreamStream } from "../helpers/fakeUpstreamStream.ts";
+import { createSSEStream } from "../../open-sse/utils/stream.ts";
+import { FORMATS } from "../../open-sse/translator/formats.ts";
+
+/** Drain a ReadableStream to a string, with optional timeout guard. */
+async function drain(out: ReadableStream, timeoutMs = 5000): Promise<string> {
+  const r = out.getReader();
+  const dec = new TextDecoder();
+  let s = "";
+  const timeout = new Promise<never>((_, reject) =>
+    setTimeout(() => reject(new Error(`drain timeout after ${timeoutMs}ms`)), timeoutMs)
+  );
+  const read = async () => {
+    for (;;) {
+      const { done, value } = await r.read();
+      if (done) break;
+      s += dec.decode(value);
+    }
+    return s;
+  };
+  return Promise.race([read(), timeout]);
+}
+
+function makeStream(extraOpts: Record<string, unknown> = {}) {
+  const up = fakeUpstreamStream();
+  const transform = createSSEStream({
+    sourceFormat: FORMATS.OPENAI,
+    targetFormat: FORMATS.OPENAI,
+    model: "m",
+    ...extraOpts,
+  });
+  const out = up.stream.pipeThrough(transform as TransformStream);
+  return { up, out };
+}
+
+test("1. stream closes after [DONE] (no hang)", async () => {
+  const { up, out } = makeStream();
+  up.push('data: {"choices":[{"delta":{"content":"hi"}}]}\n\n');
+  up.push("data: [DONE]\n\n");
+  up.close();
+  // drain() must return within the timeout — proves the stream closed
+  const text = await drain(out);
+  assert.ok(text.includes("hi"), `expected 'hi' in output: ${JSON.stringify(text)}`);
+});
+
+test("2. client cancel propagates to upstream (abort propagation)", async () => {
+  const { up, out } = makeStream();
+  let cancelled = false;
+  up.onCancel(() => {
+    cancelled = true;
+  });
+  const r = out.getReader();
+  await r.cancel("client-abort");
+  // Allow microtask queue to flush
+  await new Promise((res) => setTimeout(res, 50));
+  assert.equal(cancelled, true, "upstream cancel callback must have been called");
+});
+
+test("3. no leaked idle timers across N sequential streams", async () => {
+  // createSSEStream installs a setInterval idle watchdog per stream.
+  // If cleanup (clearInterval) does not run on stream close, timers accumulate.
+  // This test creates 10 streams and drains them; it acts as a smoke test that
+  // the process does not hang (a leaked setInterval that fires 10s later would
+  // prevent the test process from exiting cleanly in --test-force-exit mode).
+  for (let i = 0; i < 10; i++) {
+    const { up, out } = makeStream();
+    up.push("data: [DONE]\n\n");
+    up.close();
+    await drain(out);
+  }
+  // If we reach here without a timeout, no blocking resources were leaked.
+  assert.ok(true, "all 10 streams completed without hanging");
+});
+
+test("4. final snapshot does not duplicate tail text", async () => {
+  // Regression guard for the SSE snapshot bug (CLAUDE.md §2, Fase 8 B spec §4.2):
+  // the text 'Hello' should appear EXACTLY ONCE in the output, not duplicated.
+  const { up, out } = makeStream();
+  up.push('data: {"choices":[{"delta":{"content":"Hello"}}]}\n\n');
+  up.push("data: [DONE]\n\n");
+  up.close();
+  const text = await drain(out);
+  const occurrences = (text.match(/Hello/g) ?? []).length;
+  assert.equal(
+    occurrences,
+    1,
+    `'Hello' appeared ${occurrences} times; expected 1. Output: ${JSON.stringify(text)}`
+  );
+});
+
+test("5. upstream error propagates and closes stream (no hang, Hard Rule #6)", async () => {
+  // If upstream errors, the TransformStream must propagate the error so the
+  // consumer sees a rejection — never silently swallow and never hang.
+  const { up, out } = makeStream();
+  up.error(new Error("upstream boom"));
+  await assert.rejects(
+    async () => {
+      await drain(out, 2000);
+    },
+    undefined, // any error is acceptable — just must not hang
+    "upstream error must propagate as a rejection to the consumer"
+  );
+});
--- a/tests/snapshots/translation/claude-to-openai/basic-message.json
+++ b/tests/snapshots/translation/claude-to-openai/basic-message.json
@@ -0,0 +1,11 @@
+{
+  "max_tokens": 512,
+  "messages": [
+    {
+      "content": "Hello",
+      "role": "user"
+    }
+  ],
+  "model": "gpt-4o",
+  "stream": true
+}
--- a/tests/snapshots/translation/claude-to-openai/with-system.json
+++ b/tests/snapshots/translation/claude-to-openai/with-system.json
@@ -0,0 +1,15 @@
+{
+  "max_tokens": 256,
+  "messages": [
+    {
+      "content": "Be helpful.",
+      "role": "system"
+    },
+    {
+      "content": "What is 1+1?",
+      "role": "user"
+    }
+  ],
+  "model": "gpt-4o-mini",
+  "stream": true
+}
--- a/tests/snapshots/translation/gemini-to-openai/basic-chat.json
+++ b/tests/snapshots/translation/gemini-to-openai/basic-chat.json
@@ -0,0 +1,10 @@
+{
+  "messages": [
+    {
+      "content": "Hello from Gemini format!",
+      "role": "user"
+    }
+  ],
+  "model": "gpt-4o",
+  "stream": true
+}
--- a/tests/snapshots/translation/gemini-to-openai/multi-turn.json
+++ b/tests/snapshots/translation/gemini-to-openai/multi-turn.json
@@ -0,0 +1,18 @@
+{
+  "messages": [
+    {
+      "content": "What is 2+2?",
+      "role": "user"
+    },
+    {
+      "content": "4",
+      "role": "assistant"
+    },
+    {
+      "content": "And 3+3?",
+      "role": "user"
+    }
+  ],
+  "model": "gpt-4o-mini",
+  "stream": true
+}
--- a/tests/snapshots/translation/openai-to-claude/basic-chat.json
+++ b/tests/snapshots/translation/openai-to-claude/basic-chat.json
@@ -0,0 +1,16 @@
+{
+  "max_tokens": 32768,
+  "messages": [
+    {
+      "content": [
+        {
+          "text": "hi",
+          "type": "text"
+        }
+      ],
+      "role": "user"
+    }
+  ],
+  "model": "claude-opus-4-5",
+  "stream": true
+}
--- a/tests/snapshots/translation/openai-to-claude/multi-turn.json
+++ b/tests/snapshots/translation/openai-to-claude/multi-turn.json
@@ -0,0 +1,34 @@
+{
+  "max_tokens": 8192,
+  "messages": [
+    {
+      "content": [
+        {
+          "text": "What is 2+2?",
+          "type": "text"
+        }
+      ],
+      "role": "user"
+    },
+    {
+      "content": [
+        {
+          "text": "4",
+          "type": "text"
+        }
+      ],
+      "role": "assistant"
+    },
+    {
+      "content": [
+        {
+          "text": "And 3+3?",
+          "type": "text"
+        }
+      ],
+      "role": "user"
+    }
+  ],
+  "model": "claude-haiku-4-5",
+  "stream": true
+}
--- a/tests/snapshots/translation/openai-to-claude/with-system.json
+++ b/tests/snapshots/translation/openai-to-claude/with-system.json
@@ -0,0 +1,22 @@
+{
+  "max_tokens": 1024,
+  "messages": [
+    {
+      "content": [
+        {
+          "text": "Hello!",
+          "type": "text"
+        }
+      ],
+      "role": "user"
+    }
+  ],
+  "model": "claude-sonnet-4-6",
+  "stream": true,
+  "system": [
+    {
+      "text": "You are a helpful assistant.",
+      "type": "text"
+    }
+  ]
+}
--- a/tests/snapshots/translation/openai-to-gemini/basic-chat.json
+++ b/tests/snapshots/translation/openai-to-gemini/basic-chat.json
@@ -0,0 +1,38 @@
+{
+  "contents": [
+    {
+      "parts": [
+        {
+          "text": "Hello Gemini!"
+        }
+      ],
+      "role": "user"
+    }
+  ],
+  "generationConfig": {
+    "maxOutputTokens": 65536
+  },
+  "model": "gemini-2.5-flash",
+  "safetySettings": [
+    {
+      "category": "HARM_CATEGORY_HATE_SPEECH",
+      "threshold": "OFF"
+    },
+    {
+      "category": "HARM_CATEGORY_DANGEROUS_CONTENT",
+      "threshold": "OFF"
+    },
+    {
+      "category": "HARM_CATEGORY_SEXUALLY_EXPLICIT",
+      "threshold": "OFF"
+    },
+    {
+      "category": "HARM_CATEGORY_HARASSMENT",
+      "threshold": "OFF"
+    },
+    {
+      "category": "HARM_CATEGORY_CIVIC_INTEGRITY",
+      "threshold": "OFF"
+    }
+  ]
+}
--- a/tests/snapshots/translation/openai-to-gemini/with-system.json
+++ b/tests/snapshots/translation/openai-to-gemini/with-system.json
@@ -0,0 +1,47 @@
+{
+  "contents": [
+    {
+      "parts": [
+        {
+          "text": "Tell me a joke."
+        }
+      ],
+      "role": "user"
+    }
+  ],
+  "generationConfig": {
+    "maxOutputTokens": 8192,
+    "temperature": 0.7
+  },
+  "model": "gemini-2.5-pro",
+  "safetySettings": [
+    {
+      "category": "HARM_CATEGORY_HATE_SPEECH",
+      "threshold": "OFF"
+    },
+    {
+      "category": "HARM_CATEGORY_DANGEROUS_CONTENT",
+      "threshold": "OFF"
+    },
+    {
+      "category": "HARM_CATEGORY_SEXUALLY_EXPLICIT",
+      "threshold": "OFF"
+    },
+    {
+      "category": "HARM_CATEGORY_HARASSMENT",
+      "threshold": "OFF"
+    },
+    {
+      "category": "HARM_CATEGORY_CIVIC_INTEGRITY",
+      "threshold": "OFF"
+    }
+  ],
+  "systemInstruction": {
+    "parts": [
+      {
+        "text": "You are a helpful assistant."
+      }
+    ],
+    "role": "system"
+  }
+}
--- a/tests/unit/correctness/combo.property.test.ts
+++ b/tests/unit/correctness/combo.property.test.ts
@@ -0,0 +1,66 @@
+import test from "node:test";
+import assert from "node:assert/strict";
+import fc from "fast-check";
+import { configureProperties } from "../../helpers/propertyConfig.ts";
+import { validateComboDAG, resolveNestedComboModels } from "../../../open-sse/services/combo.ts";
+
+configureProperties();
+
+// ComboLike shape: { name: string, models: unknown[] }
+// normalizeModelEntry extracts .model (string) from each entry — plain strings work directly.
+// "combo:X" notation is NOT used; the resolver checks if models[i] matches combo.name in allCombos.
+
+test("validateComboDAG throws on self-cycle", () => {
+  // c0 references itself directly — must throw
+  const combos = [{ name: "c0", models: ["c0"] }];
+  assert.throws(() => validateComboDAG("c0", combos));
+});
+
+test("validateComboDAG throws on indirect cycle (a->b->a)", () => {
+  const combos = [
+    { name: "a", models: ["b"] },
+    { name: "b", models: ["a"] },
+  ];
+  assert.throws(() => validateComboDAG("a", combos));
+});
+
+test("validateComboDAG does not throw on acyclic graph", () => {
+  const combos = [
+    { name: "a", models: ["b", "openai/gpt-4o"] },
+    { name: "b", models: ["anthropic/claude-3-5-sonnet"] },
+  ];
+  assert.doesNotThrow(() => validateComboDAG("a", combos));
+});
+
+test("resolveNestedComboModels never throws and is cycle-safe", () => {
+  // Arbitrary combo graph: names c0..cN where each combo's models reference
+  // other combo names by their exact name string (no "combo:" prefix needed).
+  const graphArb = fc.integer({ min: 1, max: 5 }).chain((n) => {
+    const names = Array.from({ length: n }, (_, i) => `c${i}`);
+    const comboArr = names.map((name, i) => {
+      // Each combo references up to 2 other combos by name plus one real model
+      const refs = names.filter((_, j) => j !== i).slice(0, 2);
+      return { name, models: [...refs, "openai/gpt-4o"] };
+    });
+    return fc.constant(comboArr);
+  });
+
+  fc.assert(
+    fc.property(graphArb, (combos) => {
+      // resolveNestedComboModels must always return an array (never throw, never loop)
+      const out = resolveNestedComboModels(combos[0], combos);
+      assert.ok(Array.isArray(out), "must return array");
+    })
+  );
+});
+
+test("resolveNestedComboModels returns [] on direct cycle (cycle-safety)", () => {
+  // Cycle: c0->c1->c0
+  const combos = [
+    { name: "c0", models: ["c1"] },
+    { name: "c1", models: ["c0"] },
+  ];
+  const out = resolveNestedComboModels(combos[0], combos);
+  // When a cycle is detected, the visited guard returns [] — no infinite loop
+  assert.ok(Array.isArray(out));
+});
--- a/tests/unit/correctness/fakeUpstreamStream.test.ts
+++ b/tests/unit/correctness/fakeUpstreamStream.test.ts
@@ -0,0 +1,39 @@
+import test from "node:test";
+import assert from "node:assert/strict";
+import { fakeUpstreamStream } from "../../helpers/fakeUpstreamStream.ts";
+
+test("fakeUpstreamStream emits pushed chunks then closes", async () => {
+  const { stream, push, close } = fakeUpstreamStream();
+  push("a");
+  push("b");
+  close();
+  const reader = stream.getReader();
+  const dec = new TextDecoder();
+  let out = "";
+  for (;;) {
+    const { done, value } = await reader.read();
+    if (done) break;
+    out += dec.decode(value);
+  }
+  assert.equal(out, "ab");
+});
+
+test("fakeUpstreamStream signals cancel when consumer aborts", async () => {
+  const { stream, onCancel } = fakeUpstreamStream();
+  let cancelled = false;
+  onCancel(() => {
+    cancelled = true;
+  });
+  const reader = stream.getReader();
+  await reader.cancel("client-abort");
+  assert.equal(cancelled, true);
+});
+
+test("fakeUpstreamStream propagates error to reader", async () => {
+  const { stream, error } = fakeUpstreamStream();
+  error(new Error("upstream boom"));
+  const reader = stream.getReader();
+  await assert.rejects(async () => {
+    await reader.read();
+  }, /upstream boom/);
+});
--- a/tests/unit/correctness/fixtures.test.ts
+++ b/tests/unit/correctness/fixtures.test.ts
@@ -0,0 +1,18 @@
+import test from "node:test";
+import assert from "node:assert/strict";
+import { loadTranslationFixtures, loadSseSequences } from "../../helpers/translationFixtures.ts";
+
+test("fixtures: each pair file has at least one well-formed case", () => {
+  const cases = loadTranslationFixtures();
+  assert.ok(cases.length >= 4);
+  for (const c of cases) {
+    assert.ok(c.name && c.sourceFormat && c.targetFormat && c.input);
+  }
+});
+test("fixtures: sse sequences have chunks and expectedText", () => {
+  const seqs = loadSseSequences();
+  assert.ok(seqs.length >= 1);
+  for (const s of seqs) {
+    assert.ok(Array.isArray(s.chunks) && typeof s.expectedText === "string");
+  }
+});
--- a/tests/unit/correctness/goldenSnapshot.test.ts
+++ b/tests/unit/correctness/goldenSnapshot.test.ts
@@ -0,0 +1,41 @@
+import test from "node:test";
+import assert from "node:assert/strict";
+import fs from "node:fs";
+import os from "node:os";
+import path from "node:path";
+import { goldenSnapshot } from "../../helpers/goldenSnapshot.ts";
+
+// Use an isolated tmpdir so the selftest does not pollute tests/snapshots/
+let tmpDir: string;
+
+test("goldenSnapshot writes on first run then matches", (t) => {
+  tmpDir = fs.mkdtempSync(path.join(os.tmpdir(), "golden-selftest-"));
+
+  // First run: UPDATE_GOLDEN=1 → writes
+  process.env.UPDATE_GOLDEN = "1";
+  goldenSnapshot("selftest/sample", { b: 2, a: 1 }, tmpDir);
+  delete process.env.UPDATE_GOLDEN;
+
+  // Re-run without flag: same value (keys re-ordered) → should pass
+  assert.doesNotThrow(() => goldenSnapshot("selftest/sample", { a: 1, b: 2 }, tmpDir));
+
+  // Mismatch: different value → should throw
+  assert.throws(() => goldenSnapshot("selftest/sample", { a: 1, b: 3 }, tmpDir));
+
+  // Cleanup
+  fs.rmSync(tmpDir, { recursive: true, force: true });
+});
+
+test("goldenSnapshot first-run (no UPDATE_GOLDEN) writes and passes", () => {
+  const td = fs.mkdtempSync(path.join(os.tmpdir(), "golden-first-run-"));
+  try {
+    // File does not exist yet: should write and not throw
+    assert.doesNotThrow(() => goldenSnapshot("test/value", { x: 42 }, td));
+    // File now exists: same value should pass
+    assert.doesNotThrow(() => goldenSnapshot("test/value", { x: 42 }, td));
+    // File exists: different value should throw
+    assert.throws(() => goldenSnapshot("test/value", { x: 99 }, td));
+  } finally {
+    fs.rmSync(td, { recursive: true, force: true });
+  }
+});
--- a/tests/unit/correctness/sanitizers.property.test.ts
+++ b/tests/unit/correctness/sanitizers.property.test.ts
@@ -0,0 +1,42 @@
+import test from "node:test";
+import assert from "node:assert/strict";
+import fc from "fast-check";
+import { configureProperties } from "../../helpers/propertyConfig.ts";
+import { sanitizeErrorMessage } from "../../../open-sse/utils/error.ts";
+
+configureProperties();
+
+test("sanitizeErrorMessage never leaks a file path / stack frame", () => {
+  // sanitizeErrorMessage takes only the FIRST LINE of input (observed real behavior:
+  // it splits on \n and processes only the part before the first newline).
+  // Stack frames are on subsequent lines and thus already stripped.
+  // The invariant we test: single-line content containing "at /path/file.ts" has the
+  // absolute path replaced with "<path>", so the output never contains "at /".
+  const firstLineWithPath = fc
+    .string()
+    .map((s) => s.replace(/\n/g, " ")) // ensure single line
+    .chain((prefix) =>
+      fc
+        .string()
+        .map((s) => s.replace(/\n/g, " "))
+        .map((suffix) => `${prefix} at /home/app/open-sse/foo.ts:42:10 ${suffix}`)
+    );
+
+  fc.assert(
+    fc.property(firstLineWithPath, (input) => {
+      const out = sanitizeErrorMessage(input);
+      assert.ok(!out.includes("at /"), `leaked path in: ${JSON.stringify(out)}`);
+    })
+  );
+});
+
+test("sanitizeErrorMessage terminates on long adversarial input (ReDoS guard)", () => {
+  fc.assert(
+    fc.property(fc.integer({ min: 1000, max: 20000 }), (len) => {
+      const start = process.hrtime.bigint();
+      sanitizeErrorMessage("a".repeat(len) + "@" + "b".repeat(len) + ".com " + "1".repeat(len));
+      const ms = Number(process.hrtime.bigint() - start) / 1e6;
+      assert.ok(ms < 250, `too slow: ${ms}ms for len=${len}`);
+    })
+  );
+});
--- a/tests/unit/correctness/sse-parser.property.test.ts
+++ b/tests/unit/correctness/sse-parser.property.test.ts
@@ -0,0 +1,63 @@
+import test from "node:test";
+import assert from "node:assert/strict";
+import fc from "fast-check";
+import { configureProperties } from "../../helpers/propertyConfig.ts";
+import { parseSSELine } from "../../../open-sse/utils/streamHelpers.ts";
+import { loadSseSequences } from "../../helpers/translationFixtures.ts";
+
+configureProperties();
+
+// parseSSELine is line-oriented: it processes one line at a time.
+// The split-mid-line scenario (splitting in the middle of "data: {...}") belongs to
+// the stream-layer harness (Task 11). Here we test line-level reconstruction invariance:
+// given a complete SSE event line, parseSSELine extracts the correct content regardless
+// of what other lines surround it.
+
+function extractTextFromLines(rawContent: string): string {
+  let text = "";
+  for (const line of rawContent.split("\n")) {
+    const parsed = parseSSELine(line);
+    if (parsed && !(parsed as { done?: boolean }).done) {
+      // eslint-disable-next-line @typescript-eslint/no-explicit-any
+      const delta = (parsed as any)?.choices?.[0]?.delta?.content;
+      if (typeof delta === "string") text += delta;
+    }
+  }
+  return text;
+}
+
+test("SSE text reconstruction is invariant to chunk boundaries (line-level)", () => {
+  const seqs = loadSseSequences();
+  // Property: joining all chunks and re-splitting by line yields the same text
+  // as processing the chunks individually. This tests that parseSSELine is stateless
+  // and position-invariant (no dependency on which "chunk" a line comes from).
+  fc.assert(
+    fc.property(fc.integer({ min: 0, max: seqs.length - 1 }), (idx) => {
+      const seq = seqs[idx];
+      const full = seq.chunks.join("");
+      const reconstructed = extractTextFromLines(full);
+      assert.equal(reconstructed, seq.expectedText);
+    })
+  );
+});
+
+test("parseSSELine returns null for non-data lines", () => {
+  // Non-data lines (comment, event:, id:, retry:, empty) must return null
+  const nonDataLines = ["", ": comment", "event: update", "id: 42", "retry: 5000"];
+  for (const line of nonDataLines) {
+    assert.equal(parseSSELine(line), null, `expected null for line: ${JSON.stringify(line)}`);
+  }
+});
+
+test("parseSSELine returns {done:true} for [DONE] sentinel", () => {
+  const parsed = parseSSELine("data: [DONE]");
+  assert.deepEqual(parsed, { done: true });
+});
+
+test("parseSSELine correctly parses valid JSON data line", () => {
+  const line = 'data: {"choices":[{"delta":{"content":"hello"}}]}';
+  const parsed = parseSSELine(line) as Record<string, unknown>;
+  assert.ok(parsed && typeof parsed === "object");
+  // eslint-disable-next-line @typescript-eslint/no-explicit-any
+  assert.equal((parsed as any)?.choices?.[0]?.delta?.content, "hello");
+});
--- a/tests/unit/correctness/translation.golden.test.ts
+++ b/tests/unit/correctness/translation.golden.test.ts
@@ -0,0 +1,19 @@
+import test from "node:test";
+import { goldenSnapshot } from "../../helpers/goldenSnapshot.ts";
+import { loadTranslationFixtures } from "../../helpers/translationFixtures.ts";
+import { translateRequest } from "../../../open-sse/translator/index.ts";
+
+// Golden-file tests: freeze translateRequest output per fixture.
+// Regenerate with: UPDATE_GOLDEN=1 node --import tsx/esm --test tests/unit/correctness/translation.golden.test.ts
+for (const c of loadTranslationFixtures()) {
+  test(`golden: ${c.name}`, () => {
+    const out = translateRequest(
+      c.sourceFormat,
+      c.targetFormat,
+      (c.input as { model?: string }).model ?? "m",
+      c.input,
+      true
+    );
+    goldenSnapshot(`translation/${c.name}`, out);
+  });
+}
--- a/tests/unit/correctness/translators.property.test.ts
+++ b/tests/unit/correctness/translators.property.test.ts
@@ -0,0 +1,78 @@
+import test from "node:test";
+import assert from "node:assert/strict";
+import fc from "fast-check";
+import { configureProperties } from "../../helpers/propertyConfig.ts";
+import { translateRequest } from "../../../open-sse/translator/index.ts";
+import { FORMATS } from "../../../open-sse/translator/formats.ts";
+
+configureProperties();
+
+// Non-empty, non-whitespace content required: translator drops messages with empty/whitespace
+// content (observed real behavior — empty-content messages produce invalid Claude/Gemini requests).
+const messageArb = fc.record({
+  role: fc.constantFrom("user", "assistant", "system"),
+  content: fc.string({ minLength: 1 }).filter((s) => s.trim().length > 0),
+});
+const bodyArb = fc.record({
+  model: fc.string({ minLength: 1 }),
+  messages: fc.array(messageArb, { minLength: 1, maxLength: 6 }),
+});
+
+test("translateRequest openai->claude never throws and keeps messages", () => {
+  fc.assert(
+    fc.property(bodyArb, (body) => {
+      const out = translateRequest(
+        FORMATS.OPENAI,
+        FORMATS.CLAUDE,
+        body.model,
+        body,
+        true
+      ) as Record<string, unknown>;
+      assert.ok(out && typeof out === "object");
+      // Either messages array is non-empty, or content went to system field (system-only body)
+      const hasMessages = Array.isArray(out.messages) && (out.messages as unknown[]).length > 0;
+      const hasSystem = Array.isArray(out.system) && (out.system as unknown[]).length > 0;
+      assert.ok(hasMessages || hasSystem, "output must have messages or non-empty system");
+    })
+  );
+});
+
+test("translateRequest round-trip openai->claude->openai preserves message count for well-formed input", () => {
+  // Generate bodies with properly alternating user/assistant roles (no consecutive same role),
+  // optionally preceded by a system message. This is the well-formed case where Claude's
+  // merge-consecutive-same-role normalization does NOT collapse messages.
+  // Invariant-calibration note: arbitrary bodies with consecutive same-role messages get merged
+  // by the OpenAI→Claude translator (correct behavior — Claude doesn't allow consecutive same-role).
+  // We test only the well-formed alternating case for an exact-count invariant.
+  const altBodyArb = fc.record({
+    model: fc.string({ minLength: 1 }),
+    messages: fc.integer({ min: 1, max: 5 }).chain((n) => {
+      // Build alternating user/assistant sequence, starting with user
+      const msgs = Array.from({ length: n }, (_, i) => ({
+        role: (i % 2 === 0 ? "user" : "assistant") as "user" | "assistant",
+        content: "msg" + i,
+      }));
+      return fc.constant(msgs);
+    }),
+  });
+  fc.assert(
+    fc.property(altBodyArb, (body) => {
+      const claude = translateRequest(
+        FORMATS.OPENAI,
+        FORMATS.CLAUDE,
+        body.model,
+        body,
+        true
+      ) as Record<string, unknown>;
+      const back = translateRequest(
+        FORMATS.CLAUDE,
+        FORMATS.OPENAI,
+        body.model,
+        claude,
+        true
+      ) as Record<string, unknown>;
+      assert.ok(Array.isArray(back.messages));
+      assert.equal((back.messages as unknown[]).length, body.messages.length);
+    })
+  );
+});