Reference for creating a new deep research task via POST /v1/deepresearch/tasks.
Your Valyu API key. Get one at platform.valyu.ai
Research query or task description
"What are the key differences between RAG and fine-tuning for LLMs?"
Research mode. 'lite' is faster and good for straightforward questions. 'heavy' provides more thorough analysis for complex research.
lite, heavy "lite"
Output formats. Use ['markdown'], ['markdown', 'pdf'], or a JSON schema object for structured output. Cannot mix JSON schema with markdown/pdf.
markdown, pdf ["markdown"]Natural language strategy instructions prepended to the system prompt
"Focus on peer-reviewed sources and recent publications"
Search configuration parameters
URLs to extract content from (max 10)
10["https://example.com/article"]File attachments (PDFs, images, documents). Max 10 files.
10MCP server configurations for custom tools (max 5 servers)
5Enable/disable code execution during research
Previous deep research IDs to use as context (max 3)
3HTTPS URL for webhook notifications. When provided, Valyu will POST the full task result to this URL when the task completes or fails. The request includes X-Webhook-Signature and X-Webhook-Timestamp headers for verification. HTTP URLs are rejected—only HTTPS is supported.
Custom metadata for tracking
Task created successfully
Unique task identifier
"f992a8ab-4c91-4322-905f-190107bd5a5b"
Current task status
queued, running, completed, failed, cancelled Research mode used
lite, heavy Task creation timestamp
Custom metadata provided with the request
Whether the task is publicly accessible
32-byte cryptographic secret (64-character hex string) for verifying webhook signatures. Only returned once in the creation response—store it securely. Use this secret to verify the HMAC-SHA256 signature in the X-Webhook-Signature header of incoming webhook requests.
Additional status message (e.g., if queued)