Pinecone llms.txt

The actual file

Source Markdown

The saved llms.txt text, displayed as source. The file belongs to its publisher; the live endpoint may change.

Showing the first 18000 characters. Load the complete file to read the remainder.

# Pinecone Docs

> Official documentation for Pinecone's trusted AI knowledge platform, including Pinecone Database, Pinecone Assistant, and Pinecone Nexus.

- [Pinecone Database (550 pages)](https://docs.pinecone.io/_llms/pinecone-database.txt): Documentation for Pinecone Database.
- [Pinecone Assistant (127 pages)](https://docs.pinecone.io/_llms/pinecone-assistant.txt): Documentation for Pinecone Assistant.

## Pinecone Nexus

### Guides

#### Get started

- [Pinecone Nexus](https://docs.pinecone.io/guides/nexus/overview.md): Pinecone Nexus is the knowledge engine for agents. It compiles your data into queryable knowledge once, then serves grounded, cited answers on every call.
- [Nexus quickstart](https://docs.pinecone.io/guides/nexus/quickstart.md): Deploy Pinecone Nexus with BYOC, curate a context from your own documents, and run KnowQL queries that return grounded, cited multi-document answers.
- [Nexus key concepts](https://docs.pinecone.io/guides/nexus/concepts.md): Learn the core Pinecone Nexus concepts: sources, workspaces, contexts, manifests, artifacts, tasks, sessions, and queries, and how they connect.

#### The knowledge engine

- [How Nexus works](https://docs.pinecone.io/guides/nexus/how-it-works.md): Pinecone Nexus distills curated artifacts for retrieval, then answers with code over a retrieval SDK.
- [Nexus curation](https://docs.pinecone.io/guides/nexus/how-curation-works.md): Curation is the write path that turns sources into chunks and artifacts incrementally, driven by the manifest.
- [Nexus model guidance](https://docs.pinecone.io/guides/nexus/models.md): Learn how Pinecone Nexus uses models for generation, curation, and retrieval, and how to choose the right one.

##### Context design

- [Context design overview](https://docs.pinecone.io/guides/nexus/context-design.md): Learn how a Pinecone Nexus context's manifest turns your sources into queryable knowledge.
- [Design your own manifest](https://docs.pinecone.io/guides/nexus/design-your-own-manifest.md): Author a Pinecone Nexus context's manifest through the API by defining custom artifact and edge types, then curating.
- [Configure artifact formats](https://docs.pinecone.io/guides/nexus/configure-artifact-formats.md): Configure Pinecone Nexus artifact types as markdown prose or queryable SQLite tables through the manifest API.

##### Queries and sessions

- [Nexus queries](https://docs.pinecone.io/guides/nexus/queries.md): Pinecone Nexus queries gather evidence and compose a grounded, cited answer.
- [Set session instructions](https://docs.pinecone.io/guides/nexus/set-session-instructions.md): Set a per-session system prompt on a Pinecone Nexus query to shape how a session answers, layered over the context's default instructions.
- [Query tracing](https://docs.pinecone.io/guides/nexus/query-tracing.md): Inspect a Pinecone Nexus query trace to see reasoning steps, retrieval tool calls, token usage, latency, cost breakdown, and cache savings.

#### Tools

- [MCP server](https://docs.pinecone.io/guides/nexus/mcp-server.md): Query your Pinecone Nexus contexts from any Model Context Protocol client, such as Claude Desktop, over Streamable HTTP.

#### Bring Your Own Cloud

- [Nexus BYOC overview](https://docs.pinecone.io/guides/nexus/byoc/overview.md): Learn what Nexus BYOC is, how it relates to Database BYOC, and its architecture.
- [Nexus BYOC data residency and limits](https://docs.pinecone.io/guides/nexus/byoc/reference.md): See where your data lives and travels in Nexus BYOC, plus its authentication, encryption, cluster footprint, and limitations.
- [Deploy Nexus BYOC](https://docs.pinecone.io/guides/nexus/byoc/deploy.md): Install and operate Pinecone Nexus in your own cloud account.

### Reference

#### APIs

- [Pinecone Nexus API](https://docs.pinecone.io/reference/api/nexus/introduction.md): Programmatic access to Nexus workspaces, contexts, sources, curation, and KnowQL queries.
- [Authentication](https://docs.pinecone.io/reference/api/nexus/authentication.md): Authenticate to the Nexus API with your Pinecone API key, sent directly or exchanged for a short-lived session token.

#### Control plane

##### Manage Workspaces

- [List workspaces](https://docs.pinecone.io/api-reference/manage-workspaces/list-workspaces.md): List the workspaces in the project. Results are paginated; the `pagination` object is omitted when there are no further results.
- [Create a workspace](https://docs.pinecone.io/api-reference/manage-workspaces/create-a-workspace.md): Create a Nexus workspace. The workspace name must be unique within the project.
- [Describe a workspace](https://docs.pinecone.io/api-reference/manage-workspaces/describe-a-workspace.md): Get a description of a workspace.
- [Delete a workspace](https://docs.pinecone.io/api-reference/manage-workspaces/delete-a-workspace.md): Delete an existing workspace. Deletion is asynchronous: the workspace transitions to the `Terminating` state and its contexts are deleted, after which the workspace itself is removed.

#### Data plane

##### Auth

- [Exchange a Pinecone API key for a session token](https://docs.pinecone.io/api-reference/auth/exchange-a-pinecone-api-key-for-a-session-token.md): Exchange a Pinecone API key for the session token every other endpoint expects. Send the token as a bearer credential.
- [Current identity (whoami)](https://docs.pinecone.io/api-reference/auth/current-identity-whoami.md): Who the current credential authenticates, and which project and workspace it is scoped to.

##### Project

- [Get the active project](https://docs.pinecone.io/api-reference/project/get-the-active-project.md): The project the session is scoped to, plus whether this deployment requires a workspace host.
- [Project-wide task stats](https://docs.pinecone.io/api-reference/project/project-wide-task-stats.md): Aggregate task counters across the whole project (all contexts), broken down by state and workflow. The per-context equivalent is `GET /contexts/{slug}/tasks/stats`.

##### Connectors

- [List the project's linked connectors](https://docs.pinecone.io/api-reference/connectors/list-the-projects-linked-connectors.md): The project's linked source providers. Filter by `kind` to those valid for a context kind.
- [Get a connector](https://docs.pinecone.io/api-reference/connectors/get-a-connector.md): One linked connector, secrets stripped.
- [Unlink a connector](https://docs.pinecone.io/api-reference/connectors/unlink-a-connector.md): Unlink the connector from the project. Sources it already imported stay in place.
- [Start the OAuth link flow for a provider](https://docs.pinecone.io/api-reference/connectors/start-the-oauth-link-flow-for-a-provider.md): Returns an `authorize_url` for the user to open in a browser. The connector is persisted by the provider's OAuth redirect callback once the user approves access.
- [Link a connector with an API key](https://docs.pinecone.io/api-reference/connectors/link-a-connector-with-an-api-key.md): Link a key-based provider by supplying its API key directly, instead of running the OAuth flow.
- [Browse a connector's folders/files](https://docs.pinecone.io/api-reference/connectors/browse-a-connectors-foldersfiles.md): Browse the provider's folders and files to find the ids to import. Omit `folder` for the root.
- [List a context's connectors (with per-context enabled flag)](https://docs.pinecone.io/api-reference/connectors/list-a-contexts-connectors-with-per-context-enabled-flag.md): The project's connectors, each with whether this context may import from it.
- [Enable or disable a connector for a context](https://docs.pinecone.io/api-reference/connectors/enable-or-disable-a-connector-for-a-context.md): Enable or disable one connector for this context.

##### Contexts

- [List contexts (newest first, with stats)](https://docs.pinecone.io/api-reference/contexts/list-contexts-newest-first-with-stats.md): Every context in the active project, newest first, each with its aggregate task counters.
- [Create a context](https://docs.pinecone.io/api-reference/contexts/create-a-context.md): Created empty, and not queryable until you import sources and curate them — curate is explicit. Optionally seed a `manifest`. A `work` context is the exception: queryable from day zero, built from traces of work rather than documents.
- [Get a context](https://docs.pinecone.io/api-reference/contexts/get-a-context.md): One context and its derived lifecycle flags.
- [Update a context](https://docs.pinecone.io/api-reference/contexts/update-a-context.md): All fields optional; an absent one is left untouched. An empty string clears `description`/`guide`, and `{}` clears the manifest back to defaults.
- [Delete a context (and its Pinecone indexes)](https://docs.pinecone.io/api-reference/contexts/delete-a-context-and-its-pinecone-indexes.md): Delete the context, its sources, its curated knowledge, and its Pinecone indexes. Not reversible.
- [Per-context task statistics](https://docs.pinecone.io/api-reference/contexts/per-context-task-statistics.md): Aggregate counters over every task this context has run. The project-wide equivalent is `GET /stats`.
- [Trigger an on-demand self-tuning optimize](https://docs.pinecone.io/api-reference/contexts/trigger-an-on-demand-self-tuning-optimize.md): Tunes the manifest from real query traffic, then chains a forced re-curate. Every field of the required body is optional, so `{}` is valid — and a no-op without `candidate_queries`. The context must be curated, and only one optimize runs at a time.
- [Run the Design-flow explore agent (propose a manifest)](https://docs.pinecone.io/api-reference/contexts/run-the-design-flow-explore-agent-propose-a-manifest.md): Inspects the uploaded source and proposes manifest-template matches in the task's `output` (see `ExploreOutput`). Persists nothing — review the proposal, then apply it with a manifest update and a forced curate. Body is optional.
- [Estimate the token + time cost of curating the in-progress manifest](https://docs.pinecone.io/api-reference/contexts/estimate-the-token-+-time-cost-of-curating-the-in-progress-manifest.md): Estimates the token and time cost of curating the sources under the manifest in the request body, reported as the task's `output` (see `ProfileEstimateOutput`). Persists nothing. Body is optional.

##### Manifest

- [Get the context's manifest](https://docs.pinecone.io/api-reference/manifest/get-the-contexts-manifest.md): The context's pinned manifest document. A context that pins nothing returns `null` — the runtime fills defaults. Writes go through the `manifest` field on `PUT /contexts/{slug}`.
- [List the built-in manifest templates](https://docs.pinecone.io/api-reference/manifest/list-the-built-in-manifest-templates.md): The templates a new context can be seeded from.

##### Curation

- [Get the curation ledger](https://docs.pinecone.io/api-reference/curation/get-the-curation-ledger.md): What the last curate recorded: the per-source hash and timestamp map, artifact edges, corpus groups, the live version pin, delete tombstones, artifact-reclaim intents, and the pointer and glossary resume cursors. Chunk ids are tracked too but not served — an internal retrieval unit.
- [Build or rebuild the context's index from its sources](https://docs.pinecone.io/api-reference/curation/build-or-rebuild-the-contexts-index-from-its-sources.md): Curate the staged sources under the active manifest. **Required before querying** — there is no auto-curate. Body is optional; `force: true` rebuilds fully, which is what a manifest edit needs. Search contexts only: a work context builds via `work`/`groom`.
- [Drop named sources from the index + ledger](https://docs.pinecone.io/api-reference/curation/drop-named-sources-from-the-index-+-ledger.md): Drop named sources from the index and the curation ledger without re-reading the rest of the corpus.

##### Source Files

- [Import source documents from a project connector](https://docs.pinecone.io/api-reference/source-files/import-source-documents-from-a-project-connector.md): Pull source documents from a linked project connector (Box, ...) into the context. Supply either `folder` or `files`. Stages the source only — curate explicitly before querying.
- [Upload a source file or archive](https://docs.pinecone.io/api-reference/source-files/upload-a-source-file-or-archive.md): One file per request, max 2 GiB; archives (`.zip`, `.tar`, `.tar.gz`, `.tgz`) are expanded by the import runtime. Stages the source without indexing it — curate explicitly before querying.
- [List the source file tree (root)](https://docs.pinecone.io/api-reference/source-files/list-the-source-file-tree-root.md): The root of the source tree — what has been staged but not necessarily curated.
- [Source file/dir/size counts](https://docs.pinecone.io/api-reference/source-files/source-filedirsize-counts.md): Aggregate counts. Where source limits apply, five more fields carry the caps and usage against them (`max_files_per_context`, `max_bytes_per_context`, `max_bytes_per_file`, `used_files`, `used_bytes`); they are absent otherwise.
- [Flat manifest of source files (path, size, last_modified)](https://docs.pinecone.io/api-reference/source-files/flat-manifest-of-source-files-path-size-last_modified.md): Flat source-object inventory. Excludes the `_inbox/` staging directory.
- [List the source file tree under a path](https://docs.pinecone.io/api-reference/source-files/list-the-source-file-tree-under-a-path.md): One directory of the source tree.
- [Read a source file (raw bytes)](https://docs.pinecone.io/api-reference/source-files/read-a-source-file-raw-bytes.md): The raw bytes of one staged source file.
- [Delete a source file](https://docs.pinecone.io/api-reference/source-files/delete-a-source-file.md): Remove one file from the source tree. Its index entries survive until the next curate.

##### Knowledge

- [List curated knowledge (root)](https://docs.pinecone.io/api-reference/knowledge/list-curated-knowledge-root.md): The root of the curated knowledge tree — chunks and artifacts the last curate wrote.
- [Knowledge file/dir/size counts](https://docs.pinecone.io/api-reference/knowledge/knowledge-filedirsize-counts.md): File, directory, and byte counts over the curated knowledge tree.
- [List curated knowledge under a path](https://docs.pinecone.io/api-reference/knowledge/list-curated-knowledge-under-a-path.md): One directory of the curated knowledge tree.
- [Read a curated knowledge file (raw bytes)](https://docs.pinecone.io/api-reference/knowledge/read-a-curated-knowledge-file-raw-bytes.md): The raw bytes of one curated knowledge file.

##### Query

- [Run one KnowQL query turn](https://docs.pinecone.io/api-reference/query/run-one-knowql-query-turn.md): Send `ask` and, on a new session, a `scope` of 1–10 contexts; continue an existing one with `session_id` or `previous_query_id`. Scoped search contexts must be curated, and a scope may not mix work and search contexts. Read the answer from `output[].content[].text`.
- [Fetch one query (turn)](https://docs.pinecone.io/api-reference/query/fetch-one-query-turn.md): One turn, including its steps, citations, and usage. This is what a background run is polled with.
- [Cancel an in-progress query turn](https://docs.pinecone.io/api-reference/query/cancel-an-in-progress-query-turn.md): Idempotent — an already-terminal query is returned unchanged. On success the turn's `status` becomes `cancelled`.
- [Fetch the recorded trace for a query turn](https://docs.pinecone.io/api-reference/query/fetch-the-recorded-trace-for-a-query-turn.md): The per-turn debug trace (steps, tool calls, strategy, cost, rollup). Trace persistence is unconditional — every turn lands a trace blob — so this is available once the turn is terminal.
- [Stream a query turn's events (SSE)](https://docs.pinecone.io/api-reference/query/stream-a-query-turns-events-sse.md): The turn's events as they land — the same `type`-named events `POST /query` emits with `stream: true`, as a standalone resumable subscription. Reconnect with `Last-Event-ID` to replay from where the stream dropped.
- [Record thumbs-up/down feedback on a query turn](https://docs.pinecone.io/api-reference/query/record-thumbs-updown-feedback-on-a-query-turn.md): Record a thumbs-up or thumbs-down on a turn. The rating is returned on the turn thereafter.
- [List the selectable query models](https://docs.pinecone.io/api-reference/query/list-the-selectable-query-models.md): What the `model` field on `POST /query` draws from: the default, every catalog entry, the tier → model-id map, per-phase defaults, and the curate-capable ids. Only entries with `available: true` are selectable; the rest are rejected with `400`.
- [List project-wide query sessions (newest first)](https://docs.pinecone.io/api-reference/query/list-project-wide-query-sessions-newest-first.md): Every conversation thread in the project, newest first.
- [Get a session with its queries (conversation order)](https://docs.pinecone.io/api-reference/query/get-a-session-with-its-queries-conversation-order.md): One session with its turns in conversation order.
- [Delete a session and its queries](https://docs.pinecone.io/api-reference/query/delete-a-session-and-its-queries.md): Delete a session and every turn belonging to it.

##### Tasks

- [List tasks (project-wide, paginated)](https://docs.pinecone.io/api-reference/tasks/list-tasks-project-wide-paginated.md): Task records across the project, newest first. Filter by workflow, state, or context, and page with `limit` plus `page` or `offset`.
- [Get a task (with steps)](https://docs.pinecone.io/api-reference/tasks/get-a-task-with-steps.md): Unlike the listing, this carries the task's `steps` — every one by default, or the newest N with `steps_limit`, where `steps_total` reports the true count.
- [Cancel/terminate a task](https://docs.pinecone.io/api-reference/tasks/cancelterminate-a-task.md): Body is optional. Max body 1 MiB.

##### Task Files

- [Read a task file (raw bytes)](ht