The required status checks build_mem0 (3.10/3.11) and build_embedchain
live in ci.yml, which was path-filtered to mem0/tests/embedchain only.
PRs that touch only cli/** or docs/** could never satisfy branch
protection because the workflow never triggered.
Adds cli/**, docs/**, and .github/workflows/** to the pull_request
paths so required checks actually run on CLI and docs PRs.
Avoiding a deploy-hook secret on a public repo. Mintlify rebuilds
will continue via the GitHub App's path-detection (with the existing
"chore: trigger Mintlify redeploy" workaround when needed).
Wraps up the AGENTRUSH CLI release across both clients.
Lint:
- Node Biome: collapse multi-line printError() in whoami.ts
- Python ruff: drop unused BRAND_COLOR import; replace EN DASH with
HYPHEN-MINUS in app.py docstring
Release:
- Bump @mem0/cli and mem0-cli from 0.2.6 → 0.2.7 (patch)
- Add CHANGELOG.md (Keep a Changelog format) to both CLIs, seeded
with the 0.2.7 entry covering whoami, agent-rush <add|search>, the
PII safety prompt, and the new agent_rush.acknowledged_at config field
Docs:
- docs/platform/cli.mdx: new sections for `mem0 whoami` and
`mem0 agent-rush <add|search>` with rule table, error-code hints,
the launch copy-paste block, and the public-memory warning
- llms.txt coverage already in sync (no new .mdx files)
Mintlify auto-deploy:
- New .github/workflows/mintlify-deploy.yml — pings the deploy hook on
every push to main, bypassing the GitHub App's path-detection bug
that caused docs.mem0.ai drift (the recurring "trigger redeploy"
commits). Workflow is a clean no-op until MINTLIFY_DEPLOY_HOOK_URL
repo secret is configured in dashboard.
Adds a one-time PII warning shown before the first `mem0 agent-rush
add`. Persists acknowledgment in config so we never ask the same
machine twice.
Interactive (TTY):
⚠️ AGENTRUSH memories are PUBLIC — visible to any other player.
Do not include real names, emails, secrets, work content, or PII.
Continue? [y/N]
→ on "y" the timestamp lands in `agent_rush.acknowledged_at`.
→ anything else aborts with no add.
Non-interactive (agent invocation, no TTY): the warning prints to
stderr (visible in the agent transcript so the human notices) and
the add proceeds. Agents can't answer y/N; blocking them would defeat
the entire launch flow. Ack is intentionally not persisted in this
path so a later human run still sees the prompt.
Mirrored across Node + Python CLIs with matching JSON schema:
config.json adds `agent_rush.acknowledged_at: ""` (ISO timestamp).
Mirrors implementation across Node (TypeScript) and Python CLIs.
mem0 whoami:
Prints the agent's default_user_id from local config — the AGENTRUSH
leaderboard identifier. Used by players to cross-verify their row.
mem0 agent-rush <add|search>:
Wraps the platform's /v1/agent-rush/{memories/, memories/search/}
endpoints. Project routing is implicit (server-resolved); no flags.
Pretty-prints platform error codes into actionable hints (e.g.
agentrush_search_first → "Run 3 'mem0 agent-rush search' commands
before adding.").
Adds X-Mem0-Mode: agent-rush header for telemetry. Reuses existing
agent-mode API key from config; surfaces a graceful error if config
is missing or the key was created outside agent-mode.
Platform PR: mem0ai/platform#2804
Mgeeeek