Add coding agent timeline memory support

.gitignore

@@ -36,6 +36,34 @@ target/
 ### Mac OS ###
 .DS_Store
 
+### Python ###
+__pycache__/
+*.py[cod]
+
+### JavaScript / TypeScript ###
+node_modules/
+.pnpm-store/
+dist/
+dist-ssr/
+coverage/
+.turbo/
+.next/
+.vite/
+.vitest-attachments/
+**/__screenshots__/
+*.tsbuildinfo
+*.tgz
+npm-debug.log*
+yarn-debug.log*
+yarn-error.log*
+pnpm-debug.log*
+
+### Go ###
+*.test
+*.out
+coverage.out
+*.coverprofile
+
 .flattened-pom.xml
 **/dependency-reduced-pom.xml
 

evaluation/rawdata-agent/README.md

@@ -0,0 +1,31 @@
+# rawdata-agent Evaluation Fixtures
+
+This directory contains small black-box fixtures for validating Memind's
+`agent_timeline` ingestion path against a running `memind-server`.
+
+Each fixture is a self-contained JSON document:
+
+- `request`: payload for `POST /open/v1/memory/sync/extract`.
+- `expectations.expectedCategories`: item categories expected from the first extraction.
+- `expectations.queries`: retrieval checks for `POST /open/v1/memory/retrieve`.
+- `expectations.optionalCategories`: categories that may appear when LLM extraction is enabled,
+  but are not required for deterministic v1 acceptance.
+
+The fixtures are intentionally deterministic-friendly. They validate the core
+coding-agent memory path without depending on an LLM:
+
+- command/tool reuse memory, such as `npm test payment`.
+- resolution memory from a failed command, file edit, and later matching success.
+- duplicate submission behavior through stable raw content and item hashes.
+
+## Run
+
+Start `memind-server` on the default port, then run:
+
+```bash
+python3 evaluation/rawdata-agent/run-fixtures.py --base-url http://127.0.0.1:8366
+```
+
+The runner exits non-zero when the server is unreachable or any fixture fails.
+It prints a duplicate item rate for each fixture so regressions in idempotency
+are visible during manual evaluation.

evaluation/rawdata-agent/fixtures/auth-jwt-fix.json

@@ -0,0 +1,97 @@
+{
+  "name": "auth-jwt-fix",
+  "description": "A coding-agent episode where a JWT validation regression is fixed and validated.",
+  "request": {
+    "userId": "eval-agent-user-auth",
+    "agentId": "eval-coding-agent",
+    "sourceClient": "claude-code",
+    "rawContent": {
+      "type": "agent_timeline",
+      "sourceClient": "claude-code",
+      "sourceVersion": "eval-fixture",
+      "sessionId": "eval-auth-jwt-session",
+      "timelineId": "eval-auth-jwt-timeline",
+      "project": {
+        "name": "identity-service",
+        "rootPath": "/workspace/identity-service",
+        "git": {
+          "branch": "fix/jwt-clock-skew",
+          "commit": "8a12c0ffee"
+        }
+      },
+      "events": [
+        {
+          "id": "auth-e1",
+          "seq": 1,
+          "kind": "user_prompt",
+          "text": "Fix the expired JWT acceptance regression in the auth middleware.",
+          "occurredAt": "2026-05-24T09:00:00Z"
+        },
+        {
+          "id": "auth-e2",
+          "seq": 2,
+          "kind": "command",
+          "toolName": "Bash",
+          "command": "npm test auth-jwt",
+          "status": "failed",
+          "exitCode": 1,
+          "output": "TokenValidationError: expired JWT accepted beyond allowed clock skew",
+          "metadata": {
+            "failureSignal": "expired JWT accepted beyond allowed clock skew"
+          },
+          "occurredAt": "2026-05-24T09:01:00Z"
+        },
+        {
+          "id": "auth-e3",
+          "seq": 3,
+          "kind": "file_edit",
+          "path": "src/auth/jwt_validator.ts",
+          "operation": "modify",
+          "occurredAt": "2026-05-24T09:04:00Z"
+        },
+        {
+          "id": "auth-e4",
+          "seq": 4,
+          "kind": "file_edit",
+          "path": "test/auth/jwt_validator.test.ts",
+          "operation": "modify",
+          "occurredAt": "2026-05-24T09:06:00Z"
+        },
+        {
+          "id": "auth-e5",
+          "seq": 5,
+          "kind": "command",
+          "toolName": "Bash",
+          "command": "npm test auth-jwt",
+          "status": "success",
+          "exitCode": 0,
+          "output": "auth-jwt tests passed",
+          "occurredAt": "2026-05-24T09:08:00Z"
+        },
+        {
+          "id": "auth-e6",
+          "seq": 6,
+          "kind": "task_completed",
+          "status": "success",
+          "text": "JWT expiry validation now rejects tokens beyond the allowed clock skew.",
+          "occurredAt": "2026-05-24T09:09:00Z"
+        }
+      ]
+    }
+  },
+  "expectations": {
+    "expectedCategories": ["tool", "resolution"],
+    "queries": [
+      {
+        "query": "How should auth JWT fixes be validated?",
+        "mustContain": ["npm test auth-jwt"],
+        "expectedCategories": ["tool"]
+      },
+      {
+        "query": "What fixed expired JWT acceptance?",
+        "mustContain": ["expired JWT accepted beyond allowed clock skew"],
+        "expectedCategories": ["resolution"]
+      }
+    ]
+  }
+}

evaluation/rawdata-agent/fixtures/payment-rounding-fix.json

@@ -0,0 +1,104 @@
+{
+  "name": "payment-rounding-fix",
+  "description": "A payment rounding failure is fixed, retested, and stored as tool plus resolution memory.",
+  "request": {
+    "userId": "eval-agent-user-payment",
+    "agentId": "eval-coding-agent",
+    "sourceClient": "codex",
+    "rawContent": {
+      "type": "agent_timeline",
+      "sourceClient": "codex",
+      "sourceVersion": "eval-fixture",
+      "sessionId": "eval-payment-session",
+      "timelineId": "eval-payment-timeline",
+      "project": {
+        "name": "payments-api",
+        "rootPath": "/workspace/payments-api",
+        "git": {
+          "branch": "fix/payment-rounding",
+          "commit": "9b34fade12"
+        }
+      },
+      "events": [
+        {
+          "id": "payment-e1",
+          "seq": 1,
+          "kind": "user_prompt",
+          "text": "Fix the payment calculation rounding mismatch.",
+          "occurredAt": "2026-05-24T10:00:00Z"
+        },
+        {
+          "id": "payment-e2",
+          "seq": 2,
+          "kind": "command",
+          "toolName": "Shell",
+          "command": "npm test payment",
+          "status": "failed",
+          "exitCode": 1,
+          "output": "rounding mismatch on half-cent tax calculation",
+          "metadata": {
+            "failureSignal": "rounding mismatch on half-cent tax calculation"
+          },
+          "occurredAt": "2026-05-24T10:01:00Z"
+        },
+        {
+          "id": "payment-e3",
+          "seq": 3,
+          "kind": "file_read",
+          "path": "src/payment/calc.ts",
+          "operation": "read",
+          "occurredAt": "2026-05-24T10:02:00Z"
+        },
+        {
+          "id": "payment-e4",
+          "seq": 4,
+          "kind": "file_edit",
+          "path": "src/payment/calc.ts",
+          "operation": "modify",
+          "occurredAt": "2026-05-24T10:03:00Z"
+        },
+        {
+          "id": "payment-e5",
+          "seq": 5,
+          "kind": "file_edit",
+          "path": "test/payment/calc.test.ts",
+          "operation": "modify",
+          "occurredAt": "2026-05-24T10:05:00Z"
+        },
+        {
+          "id": "payment-e6",
+          "seq": 6,
+          "kind": "command",
+          "toolName": "Shell",
+          "command": "npm test payment",
+          "status": "success",
+          "exitCode": 0,
+          "output": "payment tests passed",
+          "occurredAt": "2026-05-24T10:07:00Z"
+        },
+        {
+          "id": "payment-e7",
+          "seq": 7,
+          "kind": "stop",
+          "status": "success",
+          "occurredAt": "2026-05-24T10:08:00Z"
+        }
+      ]
+    }
+  },
+  "expectations": {
+    "expectedCategories": ["tool", "resolution"],
+    "queries": [
+      {
+        "query": "How do I validate payment calculation changes?",
+        "mustContain": ["npm test payment"],
+        "expectedCategories": ["tool"]
+      },
+      {
+        "query": "What resolved the payment rounding mismatch?",
+        "mustContain": ["rounding mismatch on half-cent tax calculation"],
+        "expectedCategories": ["resolution"]
+      }
+    ]
+  }
+}

evaluation/rawdata-agent/fixtures/project-directive.json

@@ -0,0 +1,71 @@
+{
+  "name": "project-directive",
+  "description": "A project-scoped coding-agent instruction that should be recoverable as agent memory.",
+  "request": {
+    "userId": "eval-agent-user-directive",
+    "agentId": "eval-coding-agent",
+    "sourceClient": "claude-code",
+    "rawContent": {
+      "type": "agent_timeline",
+      "sourceClient": "claude-code",
+      "sourceVersion": "eval-fixture",
+      "sessionId": "eval-directive-session",
+      "timelineId": "eval-directive-timeline",
+      "project": {
+        "name": "ledger-service",
+        "rootPath": "/workspace/ledger-service",
+        "git": {
+          "branch": "main",
+          "commit": "7c77directive"
+        }
+      },
+      "events": [
+        {
+          "id": "directive-e1",
+          "seq": 1,
+          "kind": "user_prompt",
+          "text": "For this repository, always run go test ./... before reporting completion.",
+          "occurredAt": "2026-05-24T11:00:00Z"
+        },
+        {
+          "id": "directive-e2",
+          "seq": 2,
+          "kind": "file_edit",
+          "path": "README.md",
+          "operation": "modify",
+          "occurredAt": "2026-05-24T11:01:00Z"
+        },
+        {
+          "id": "directive-e3",
+          "seq": 3,
+          "kind": "command",
+          "toolName": "Bash",
+          "command": "go test ./...",
+          "status": "success",
+          "exitCode": 0,
+          "output": "ok ./...",
+          "occurredAt": "2026-05-24T11:03:00Z"
+        },
+        {
+          "id": "directive-e4",
+          "seq": 4,
+          "kind": "task_completed",
+          "status": "success",
+          "text": "Updated README and validated with go test ./...",
+          "occurredAt": "2026-05-24T11:04:00Z"
+        }
+      ]
+    }
+  },
+  "expectations": {
+    "expectedCategories": ["tool"],
+    "queries": [
+      {
+        "query": "What command should be run before reporting completion in ledger-service?",
+        "mustContain": ["go test ./..."],
+        "expectedCategories": ["tool"]
+      }
+    ],
+    "optionalCategories": ["directive"]
+  }
+}