edith-senso-ingest
# Senso Content Ingestion
Ingest documents and content into the user's Senso.ai knowledge base so it becomes searchable through Edith smart glasses.
## When to use
Activate this skill when the user wants to add, upload, or ingest content into their knowledge base. Examples:
- "Add this to my knowledge base..."
- "Ingest this document..."
- "Store this information..."
- "Remember this for later..." (when referring to document-scale content, not short notes)
- "Upload this to Senso..."
- "Index this content..."
Do NOT use this skill for simple note-taking or reminders. This is for ingesting substantial content that should be searchable later.
## Setup
The user must have a Senso.ai API key configured. If not, tell them:
1. Sign up at https://senso.ai and create a project
2. Go to Settings > API Keys and generate a new key
3. Tell OpenClaw: "My Senso API key is sk-..." and store it for future use
The API key should be stored in OpenClaw's memory/config as `SENSO_API_KEY`.
## How to ingest content
Use the `exec` tool to call the Senso.ai content ingestion endpoint:
```bash
curl -s -X POST "https://sdk.senso.ai/api/v1/content/raw" \
-H "Content-Type: application/json" \
-H "X-API-Key: ${SENSO_API_KEY}" \
-d '{"content": "<text content to ingest>", "metadata": {"title": "<optional title>", "source": "<optional source>"}}'
```
### Ingesting from a file
If the user provides a file path, read it first and then send the content:
```bash
curl -s -X POST "https://sdk.senso.ai/api/v1/content/raw" \
-H "Content-Type: application/json" \
-H "X-API-Key: ${SENSO_API_KEY}" \
-d @- <<'BODY'
{
"content": "<file contents>",
"metadata": {
"title": "<filename>",
"source": "file"
}
}
BODY
```
For large files, consider chunking the content into logical sections before ingesting.
## Formatting responses for Edith voice output
Edith speaks responses through smart glasses speakers. Keep confirmations brief:
1. **Confirm success simply.** Say "Done, I've added that to your knowledge base" or "Got it, that's been indexed."
2. **Mention what was ingested.** Say "I've added the return policy document to your knowledge base" so the user knows what happened.
3. **No technical details.** Don't mention API responses, document IDs, chunk counts, or byte sizes.
4. **If it fails, be clear.** Say what went wrong in plain language.
### Example
User: "Hey Edith, add this meeting summary to my knowledge base: We decided to launch the new product line in Q3 and increase the marketing budget by 20 percent."
Good response: "Done, I've added your meeting summary to your knowledge base."
Bad response: "I have successfully ingested 1 document containing 147 characters into your Senso.ai knowledge base. The document ID is doc_abc123 and it was chunked into 1 segment."
## Error handling
- **Missing API key**: "You haven't set up your Senso knowledge base yet. You'll need a Senso API key. Visit senso.ai to get one, then tell me the key."
- **401 Unauthorized**: "Your Senso API key seems invalid. Please check it and try again."
- **Content too large**: Split the content into smaller pieces and ingest each separately. Confirm with "I've added your document in multiple parts to your knowledge base."
- **Network/timeout error**: "I couldn't reach your knowledge base right now. Try again in a moment."
- **Empty content**: "There's no content to add. Tell me what you'd like to store in your knowledge base."
标签
skill
ai
