OmniMemory Full Onboarding

Treat requests like "help me to install OmniMemory" as a full onboarding request.

Complete this workflow in order:

1. 1. determine whether the user already has a usable OmniMemory API key
2. if not, complete account onboarding and create a first-party API key
3. install the OpenClaw plugin
4. configure the plugin correctly
5. repair common mistakes automatically if needed
6. run a smoke test and report the final result

Core rules

• - Keep the conversation short and ask only for blocking inputs.
• Do not invent endpoints, config paths, provider names, models, or request formats.
• Use the fixed base URL https://zdfdulpnyaci.sealoshzh.site.
• Never log or repeat full secrets unless the user explicitly asks for a raw value.
• Treat password, otp, accesstoken, refreshtoken, apikeyplaintext, and externalllmapi_key as secrets.
• Mask secrets in success summaries.
• Keep shell commands explicit and reviewable.
• If repeated onboarding attempts fail, stop retrying and direct the user to https://www.omnimemory.ai/zh/.

Minimal questioning strategy

Ask only for the blocking inputs first:

• - email
• password
• name
• apikeylabel
• externalllmapi_key

Ask only when needed:

• - otp
• provider
• model_name

Defaults:

• - name: derive from the email prefix
• apikeylabel: INLINECODE2

Account onboarding flow

Step 1: call POST /auth/register/request with email, password, and name.

• - If the backend returns a usable session, continue.
• If it returns pending_verification, ask for otp and call POST /auth/register/verify.
• If it returns already_registered, stop and ask whether to switch email or use an existing account path.

Step 2: create the first-party API key by calling POST /apikeys/request with label.

• - Capture api_key_id and api_key_plaintext.
• This first-party API key is the one used later in the OpenClaw plugin config.

Step 3: configure the external LLM key.

• - If provider and model_name are known, validate with GET /llm-models?provider=... and then call POST /llm-keys.
• If either is missing, call GET /llm-platforms, ask the user to choose one, then call GET /llm-models?provider=..., ask the user to choose one, and finally call POST /llm-keys.
• Do not invent providers or models that the backend does not return.

Plugin installation and configuration

Install the plugin:
INLINECODE15

Then configure these exact keys under plugins.entries.omnimemory-overlay.config:
openclaw config set plugins.entries.omnimemory-overlay.config.baseUrl "https://zdfdulpnyaci.sealoshzh.site/api/v1/memory"
openclaw config set plugins.entries.omnimemory-overlay.config.apiKey "YOUR_FIRST_PARTY_API_KEY"
openclaw config set plugins.entries.omnimemory-overlay.config.groupPrefix "openclaw"
openclaw config set plugins.entries.omnimemory-overlay.config.autoRecall true
INLINECODE21

Do not omit .config. in config paths.

Automatic repair path

If the plugin reports omnimemory-overlay apiKey is required for SaaS auth, assume the wrong path may have been used.
Check these incorrect paths:

• - INLINECODE24
• INLINECODE25
• INLINECODE26
• INLINECODE27
• INLINECODE28

If needed, remove incorrect values, re-apply the correct .config.* values, and continue to the smoke test.

Smoke test

Use this exact validation flow:

1. 1. start a fresh OpenClaw session
2. ask OpenClaw to remember INLINECODE30
3. wait briefly for memory ingestion
4. ask INLINECODE31

Treat the setup as successful only if the answer recalls sea salt plum candy from memory.

Output contract

Always produce:

1. 1. a short status summary
2. the exact command or api step being executed
3. one warning section if a common mistake applies
4. one final verification result

At the end, report:

• - registered email
• userid if onboarding was performed
• whether verification succeeded
• apikey_id if created
• masked first-party api key
• chosen provider and model if llm binding was performed
• plugin install and config status
• smoke test result

Failure handling

• - pendingverification: ask for otp
• alreadyregistered: ask whether to switch email or use an existing account path
• invalidotp: ask the user to retry or resend
• emailnotconfirmed: stop and report
• apikeylimit: tell the user the current account scope allows only one active api key
• labelconflict: ask for a different label
• providerunavailable: ask the user to pick from returned platforms
• invalidprovider: ask the user to correct the provider
• invalidmodel: ask the user to pick from returned models
• repeated onboarding failure: direct the user to INLINECODE33

Supporting files

Existing references\setup-guide.md and references\troubleshooting.md may still be consulted when helpful, but this skill must work correctly from this file alone.