Perplexity AI API

Generate a chat completion response for the given conversation.

Search the web and retrieve relevant web page contents.

Generate embeddings for a list of texts. Use these embeddings for semantic search, clustering, and other machine learning applications.

Generate contextualized embeddings for document chunks. Chunks from the same document share context awareness, improving retrieval quality for document-based applications.

Retrieve the response for a given asynchronous chat completion request.

Retrieve a list of all asynchronous chat completion requests for a given user.

Submit an asynchronous chat completion request.

Generate a response for the provided input with optional web search and reasoning.

Retrieve a response created earlier. Returns a JSON snapshot of the response. Only responses created with `store` omitted or `true` can be retrieved; `store: false` responses return a 404.

List the files a response produced in the sandbox and delivered with the `share_file` tool. Scoped to the authenticated account, which only ever sees its own files. File content is not returned inline; download it with the file content endpoint.

Download the raw bytes of a file the response delivered with the `share_file` tool. The `file_id` comes from the file list endpoint and is distinct from the response `id`. The response streams raw bytes, not JSON, with a `Content-Type` matching the file and a `Content-Disposition: attachment` header carrying the original filename.

Requests cancellation of response `{id}`. Acts on durable state, so it works for background runs that outlive the client connection. Cancellation is asynchronous: a `200` acknowledges the request (`status: cancelling`) and the run stops shortly after - poll `GET /v1/agent/{id}` for its terminal status. Cancelling a run that is already terminal returns `400`. Ownership is enforced server side; a cross-tenant or unknown id surfaces as `404`.

List the models available for the Agent API. Returns model identifiers that can be used with the `POST /v1/agent` endpoint. The response follows the OpenAI List Models format for compatibility with third-party tools.

Returns your organization's Perplexity usage analytics as a bucketed time series for a single dataset. Requires an organization analytics API key, generated by an org admin from Settings → Organization → Computer in the Perplexity web app. Only complete UTC grid buckets are returned: start_time snaps down to its bucket, and the in-progress bucket is never included. Datasets differ in what they count: credits for credit_usage, queries for query_volume, distinct active members for daily_active_users, and thread counts for the rest. Buckets of daily_active_users are not additive — a member active on several days is counted once per day, so summing them is not the window's active-member count.

Returns your organization's Perplexity usage broken down per member, as daily buckets. Each day lists one row per member with usage that day, keyed by email. credit_usage carries the same Model and Credit Source breakdown as v1; query_volume carries Feature, Project and Comet. Uses the same organization analytics API key as v1. Only complete UTC days are returned; the in-progress day is never included.

Lists the skills in your Project, newest first.

Creates a skill from a zip bundle. The bundle must contain a `SKILL.md` that provides the skill name and description; those fields are read from the bundle, not from the request.

Returns a skill. Without `revision`, returns the active revision.

Uploads a new zip bundle as a new revision of an existing skill. Pass the revision you expect to replace as the required `expected_revision`; the request fails with `409` if the skill has already moved past it, so the update cannot race a concurrent write. Name and description are re-read from the new bundle.

Deletes a skill. Pass the current revision as `expected_revision`; the delete fails with `409` if the skill has a newer revision, so a delete never races a concurrent update.

Lists the revisions of a skill, newest first.

Returns a short-lived URL to download the skill bundle as a zip. Without `revision`, returns the active revision.