Saket Aryan
df70d7833f
Six defects in 0.3.x plugin telemetry. Defects 1, 2 and 6 were not three bugs: they were one spool protocol getting three properties wrong. Identity was decided by the wrong process. `harness` was stamped in record(), correctly, but `source` was read from a module global in flush() — so whichever process drained the spool named every event in it. Two processes never call init(): mcp_server.py, and the detached `python3 telemetry.py` sender that spawn_flush() starts. record() now stamps source beside harness, and the build generates core/_harness_id.py per host so identity resolves with no init() call at all. That also unifies two defaults that disagreed (`<host>_plugin` vs `MEM0_<HOST>_PLUGIN`), which could yield three source values for one plugin. Ownership was inferred, not held. Path.replace is os.rename, which preserves mtime, so a claim made after a quiet minute inherited the spool's age and was stealable the instant it existed. Claims are touched on creation and the per-batch rewrite doubles as a lease heartbeat. Progress was not durable. flush() returned on the first failed batch without truncating, so the retry re-posted from index 0 — 150 events delivered 250 times. It now rewrites the claim with the unsent remainder after every batch, bounding a crash to one repeated batch, and each event carries a uuid. Parked batches starved. They were only reachable when no spool existed, and because sessions keep recording there usually was one, so a batch parked by a failed send waited until the 7-day expiry deleted it unsent — despite its own presence being what starts the sender. flush() drains them in the same run, and expiry now applies only after a genuine retry has failed. code.install counted upgrades and repeat sessions. is_first_run() read the identity file, which only a successful flush writes, so an offline user recorded an install every session forever. A dedicated install-state.json is claimed atomically at record time; a non-empty data directory reads as an upgrade. The docs called this anonymous. Every event carries the account email, and the hashes were unsalted SHA-256 over a git remote URL or an absolute path containing the username. READMEs, the module docstring and a new docs section now say what the code does, and repo/session digests are salted per install. A cached email outlived an API key change. It is now re-resolved when the key's fingerprint differs, and $identify aliases anonymous->email only — aliasing one account to another merges person profiles irreversibly. All six shipped green because the shared core's only tests lived under one host, behind a conftest that calls init() at import. Core behaviour was never exercised uninitialised. Adds agent-plugin-core/tests with no init, including subprocess tests and coverage for the portable plugin, which has no flush worker and would pass a native-only test vacuously. Also puts the three surface headers on the SDKs, CLIs and integrations, and corrects a README claiming ZAPIER/STRANDS were already in the platform allowlist. Verified: 59 core tests, 203 claude-code, 11 cursor, 5 codex, 2 kimi, 6 antigravity. ruff and compileall clean. --check clean for all six hosts. TypeScript changes are not typechecked locally (deps not installed). Claude-Session: https://claude.ai/code/session_01C7tEmH86HAr7GoAAKCEHZb
96 lines
5.1 KiB
Markdown
96 lines
5.1 KiB
Markdown
# deepseek-plugin
|
|
|
|
[Mem0](https://mem0.ai) long-term memory as a native [DeepSeek Harness](https://github.com/deepseek-ai/deepseek-harness) (Cordis) plugin.
|
|
|
|
It gives a Harness agent automatic long-term memory plus two explicit memory tools backed by the Mem0 SDK:
|
|
|
|
| Capability | Does |
|
|
|---|---|
|
|
| Auto-recall | Searches Mem0 for the latest human prompt and adds unseen results to the model context |
|
|
| Auto-capture | Stores the human/assistant messages from each completed turn |
|
|
| `search_memory` | Recall facts from Mem0 relevant to a query |
|
|
| `add_memory` | Store a fact in Mem0 for future sessions |
|
|
|
|
Unlike the local/file-based memory plugins in the ecosystem, Mem0 is a managed backend: server-side extraction, semantic dedup and conflict resolution, and memories that other agents can retrieve when their user and entity filters match.
|
|
|
|
Current package version: `0.3.0`.
|
|
|
|
Sidekick is available only in the [Claude Code plugin](../claude-code-plugin/README.md#sonnet-sidekick-agent).
|
|
|
|
## How it works
|
|
|
|
A Cordis plugin is a module exporting `apply(ctx, config)`. This one waits for the Harness tool and system-prompt services, then uses the native extension points:
|
|
|
|
- `system-prompt/assemble` recalls memory before a model request.
|
|
- `session/event` captures only completed turns from the durable event stream.
|
|
- `ctx.tools.register(...)` exposes explicit search and add tools.
|
|
|
|
Completed human and assistant text is preserved after secret redaction, without the former 6,000-character per-message cutoff. Recall queries and displayed tool results retain separate size limits. These behaviors use [agent-plugin-core](../agent-plugin-core/README.md); this integration keeps its native tools and user-based scoping.
|
|
|
|
Cordis owns listener and tool cleanup when the plugin unmounts. Every automatic path is fail-open: a memory API failure does not block the agent.
|
|
|
|
```
|
|
[ mem0ai SDK ] <-- managed memory, owned by Mem0
|
|
|
|
|
[ deepseek-plugin: prompt + session listeners, memory tools ] <-- this package
|
|
|
|
|
[ DeepSeek Harness ] <-- the agent, loaded via cordis.yml
|
|
```
|
|
|
|
## Try it locally
|
|
|
|
1. Build and pack the plugin:
|
|
```sh
|
|
cd integrations/deepseek-plugin
|
|
pnpm install --frozen-lockfile
|
|
pnpm build
|
|
mkdir -p /tmp/mem0-deepseek-plugin
|
|
pnpm pack --pack-destination /tmp/mem0-deepseek-plugin
|
|
```
|
|
2. Set your Mem0 key:
|
|
```sh
|
|
export MEM0_API_KEY=...
|
|
```
|
|
3. Install it into a disposable Harness profile:
|
|
```sh
|
|
DSH_HOME=/tmp/mem0-dsh-dev pnpm dlx @deepseek-ai/dsh@0.1.1-rc.2 \
|
|
plugin --profile headless add /tmp/mem0-deepseek-plugin/mem0-deepseek-plugin-0.3.0.tgz
|
|
```
|
|
4. Copy `cordis.example.yml`, set its installed package path and your `userId`, then run Harness with the same profile:
|
|
```sh
|
|
DSH_HOME=/tmp/mem0-dsh-dev pnpm dlx @deepseek-ai/dsh@0.1.1-rc.2 \
|
|
web --patch ./integrations/deepseek-plugin/cordis.example.yml
|
|
```
|
|
5. Open http://127.0.0.1:3080 and ask the agent to remember something, then recall it in a later turn.
|
|
|
|
For a Mem0 Platform on-prem or dedicated deployment, point `config.host` at that base URL (defaults to `api.mem0.ai`). `host` overrides the Platform base URL. It does not support the self-hosted Mem0 OSS API.
|
|
|
|
## Configuration
|
|
|
|
| Field | Required | Default | Notes |
|
|
|---|---|---|---|
|
|
| `apiKey` | no | `$MEM0_API_KEY` | Mem0 platform API key |
|
|
| `userId` | yes | | Entity that owns the memories |
|
|
| `allowUserOverride` | no | `false` | Permit model-selected access to a different user only in a trusted multi-user deployment |
|
|
| `host` | no | `api.mem0.ai` | Platform base URL (on-prem / dedicated) |
|
|
| `autoRecall` | no | `true` | Recall relevant memory before model requests |
|
|
| `autoCapture` | no | `true` | Store completed human/assistant turns |
|
|
|
|
## Memory scope
|
|
|
|
Automatic capture and recall use the configured `userId` across sessions. Automatic writes do not attach a repository ID or `runId`.
|
|
|
|
Both `search_memory` and `add_memory` accept optional `agentId` and `runId`. On search, these narrow the returned memories; on add, they attach those identities to the stored memory. Pass a known `runId` to search memories explicitly saved with that session ID. This does not include automatically captured user-only memories or identify the session making the request.
|
|
|
|
Per-call `userId` overrides are rejected unless the operator enables `allowUserOverride: true`. Automatic recall and capture always use the configured user.
|
|
|
|
## Telemetry
|
|
|
|
Writes are tagged `source="DEEPSEEK_HARNESS"`, which the Mem0 backend recognizes so usage surfaces by name rather than bucketing into `OTHERS`.
|
|
|
|
The plugin also sends usage events (which tool ran, duration, result counts, coarse failure kind) so Mem0 can tell how the plugin is used and where it breaks. These are **not anonymous**: when an API key is configured they are sent under your Mem0 account email, the same way the SDK attributes its own. Queries, memory text, and entity ids are never sent. Turn it off with `MEM0_TELEMETRY=false`.
|
|
|
|
## Status
|
|
|
|
Developer preview. Tracks the DeepSeek Harness v0.1 plugin API, which is young and moving. Harness capability packages are peer dependencies supplied by the host; this package pins matching release-candidate versions for local typechecking and tests.
|