Asynchronous document parsing that converts PDFs, Office files, and images into AI-ready markdown with inferred hierarchy and positional metadata; submit a job, then poll status and results (basic text or standard multimodal).
openapi: 3.0.1
info:
title: Contextual AI Platform API
description: >-
REST API for the Contextual AI enterprise RAG platform. Provides agents
(create / query grounded RAG agents), datastores and documents (ingest and
manage the knowledge corpus), and standalone component APIs - Generate
(grounded generation with the GLM), Rerank (instruction-following reranker),
Parse (document parsing into AI-ready markdown), and LMUnit (natural-language
unit-test evaluation). All endpoints authenticate with a Bearer API key.
termsOfService: https://contextual.ai/terms-of-service/
contact:
name: Contextual AI Support
url: https://docs.contextual.ai
email: support@contextual.ai
version: '1.0'
servers:
- url: https://api.contextual.ai/v1
description: Contextual AI production API
security:
- bearerAuth: []
tags:
- name: Agents
description: Create, configure, and manage RAG agents.
- name: Agents Query
description: Query agents for grounded responses, retrievals, metrics, and feedback.
- name: Datastores
description: Create and manage datastores that hold ingested knowledge.
- name: Documents
description: Ingest and manage documents within a datastore.
- name: Generate
description: Grounded generation with the Grounded Language Model (GLM).
- name: Rerank
description: Instruction-following reranking of retrieved passages.
- name: Parse
description: Parse documents into structured, AI-ready markdown.
- name: LMUnit
description: Evaluate model responses with natural-language unit tests.
- name: Users
description: Manage workspace users.
paths:
/agents:
post:
operationId: createAgent
tags:
- Agents
summary: Create Agent
description: >-
Create a new RAG agent. If no datastore_ids are supplied, a new
datastore is created automatically and associated with the agent.
requestBody:
required: true
content:
application/json:
schema:
$ref: '#/components/schemas/CreateAgentRequest'
responses:
'200':
description: Agent created.
content:
application/json:
schema:
$ref: '#/components/schemas/CreateAgentResponse'
'422':
$ref: '#/components/responses/ValidationError'
get:
operationId: listAgents
tags:
- Agents
summary: List Agents
description: Retrieve a list of agents in the workspace, with cursor-based pagination.
parameters:
- name: cursor
in: query
required: false
schema:
type: string
description: Cursor from a previous list response for pagination.
- name: limit
in: query
required: false
schema:
type: integer
minimum: 1
maximum: 1000
description: Maximum number of agents to return.
responses:
'200':
description: A list of agents.
content:
application/json:
schema:
$ref: '#/components/schemas/ListAgentsResponse'
'422':
$ref: '#/components/responses/ValidationError'
/agents/{agent_id}:
get:
operationId: getAgentMetadata
tags:
- Agents
summary: Get Agent Metadata
parameters:
- $ref: '#/components/parameters/AgentId'
responses:
'200':
description: Agent metadata.
content:
application/json:
schema:
$ref: '#/components/schemas/Agent'
'422':
$ref: '#/components/responses/ValidationError'
patch:
operationId: editAgent
tags:
- Agents
summary: Edit Agent
parameters:
- $ref: '#/components/parameters/AgentId'
requestBody:
required: true
content:
application/json:
schema:
$ref: '#/components/schemas/EditAgentRequest'
responses:
'200':
description: Agent updated.
'422':
$ref: '#/components/responses/ValidationError'
delete:
operationId: deleteAgent
tags:
- Agents
summary: Delete Agent
parameters:
- $ref: '#/components/parameters/AgentId'
responses:
'200':
description: Agent deleted.
'422':
$ref: '#/components/responses/ValidationError'
/agents/{agent_id}/query:
post:
operationId: queryAgent
tags:
- Agents Query
summary: Query Agent
description: >-
Send a conversation to an agent and receive a grounded response with
attributions, retrieval contents, and groundedness scores. Set
stream=true to receive a Server-Sent Events token stream.
parameters:
- $ref: '#/components/parameters/AgentId'
- name: retrievals_only
in: query
required: false
schema:
type: boolean
default: false
description: Fetch retrieval content and metadata only, skipping generation.
- name: include_retrieval_content_text
in: query
required: false
schema:
type: boolean
default: false
description: Include the text of retrieved contents in the response.
requestBody:
required: true
content:
application/json:
schema:
$ref: '#/components/schemas/QueryRequest'
responses:
'200':
description: Grounded agent response.
content:
application/json:
schema:
$ref: '#/components/schemas/QueryResponse'
text/event-stream:
schema:
type: string
description: SSE token stream emitted when stream=true.
'422':
$ref: '#/components/responses/ValidationError'
/agents/{agent_id}/feedback:
post:
operationId: provideFeedback
tags:
- Agents Query
summary: Provide Feedback
description: Provide thumbs-up / thumbs-down or score feedback on a message.
parameters:
- $ref: '#/components/parameters/AgentId'
requestBody:
required: true
content:
application/json:
schema:
$ref: '#/components/schemas/FeedbackRequest'
responses:
'200':
description: Feedback recorded.
'422':
$ref: '#/components/responses/ValidationError'
/datastores:
post:
operationId: createDatastore
tags:
- Datastores
summary: Create Datastore
requestBody:
required: true
content:
application/json:
schema:
$ref: '#/components/schemas/CreateDatastoreRequest'
responses:
'200':
description: Datastore created.
content:
application/json:
schema:
$ref: '#/components/schemas/CreateDatastoreResponse'
'422':
$ref: '#/components/responses/ValidationError'
get:
operationId: listDatastores
tags:
- Datastores
summary: List Datastores
description: Retrieve a list of datastores, with cursor-based pagination.
parameters:
- name: cursor
in: query
required: false
schema:
type: string
- name: limit
in: query
required: false
schema:
type: integer
minimum: 1
maximum: 1000
responses:
'200':
description: A list of datastores.
content:
application/json:
schema:
$ref: '#/components/schemas/ListDatastoresResponse'
'422':
$ref: '#/components/responses/ValidationError'
/datastores/{datastore_id}:
get:
operationId: getDatastoreMetadata
tags:
- Datastores
summary: Get Datastore Metadata
parameters:
- $ref: '#/components/parameters/DatastoreId'
responses:
'200':
description: Datastore metadata.
content:
application/json:
schema:
$ref: '#/components/schemas/Datastore'
'422':
$ref: '#/components/responses/ValidationError'
delete:
operationId: deleteDatastore
tags:
- Datastores
summary: Delete Datastore
parameters:
- $ref: '#/components/parameters/DatastoreId'
responses:
'200':
description: Datastore deleted.
'422':
$ref: '#/components/responses/ValidationError'
/datastores/{datastore_id}/documents:
post:
operationId: ingestDocument
tags:
- Documents
summary: Ingest Document
description: >-
Ingest a document (PDF, HTML, DOC(X), PPT(X), PNG, JPG/JPEG) into a
datastore. The document is parsed and chunked for retrieval.
parameters:
- $ref: '#/components/parameters/DatastoreId'
requestBody:
required: true
content:
multipart/form-data:
schema:
type: object
required:
- file
properties:
file:
type: string
format: binary
description: Binary file to ingest.
metadata:
type: string
description: >-
Stringified JSON with a custom_metadata object (max 15
fields, <= 2 KB encoded).
configuration:
type: string
description: Stringified JSON overriding datastore settings for this document.
responses:
'200':
description: Document accepted for ingestion.
content:
application/json:
schema:
$ref: '#/components/schemas/IngestDocumentResponse'
'422':
$ref: '#/components/responses/ValidationError'
get:
operationId: listDocuments
tags:
- Documents
summary: List Documents
parameters:
- $ref: '#/components/parameters/DatastoreId'
- name: cursor
in: query
required: false
schema:
type: string
- name: limit
in: query
required: false
schema:
type: integer
responses:
'200':
description: A list of documents.
'422':
$ref: '#/components/responses/ValidationError'
/datastores/{datastore_id}/documents/{document_id}:
delete:
operationId: deleteDocument
tags:
- Documents
summary: Delete Document
parameters:
- $ref: '#/components/parameters/DatastoreId'
- name: document_id
in: path
required: true
schema:
type: string
format: uuid
responses:
'200':
description: Document deleted.
'422':
$ref: '#/components/responses/ValidationError'
/generate:
post:
operationId: generate
tags:
- Generate
summary: Generate
description: >-
Generate a response grounded in supplied knowledge using the Grounded
Language Model (GLM). The total request is limited to 32,000 tokens.
requestBody:
required: true
content:
application/json:
schema:
$ref: '#/components/schemas/GenerateRequest'
responses:
'200':
description: Grounded generation result.
content:
application/json:
schema:
$ref: '#/components/schemas/GenerateResponse'
'422':
$ref: '#/components/responses/ValidationError'
/rerank:
post:
operationId: rerank
tags:
- Rerank
summary: Rerank
description: >-
Rerank a set of documents against a query using the instruction-following
reranker. Optionally steer ranking with a natural-language instruction.
requestBody:
required: true
content:
application/json:
schema:
$ref: '#/components/schemas/RerankRequest'
responses:
'200':
description: Reranked results.
content:
application/json:
schema:
$ref: '#/components/schemas/RerankResponse'
'422':
$ref: '#/components/responses/ValidationError'
/parse:
post:
operationId: parseFile
tags:
- Parse
summary: Parse File
description: >-
Submit a document (PDF, DOC(X), PPT(X), PNG, JPG/JPEG; max 300 MB,
2000 pages) to be parsed into AI-ready markdown. Returns a job_id;
poll the status and results endpoints. Results are retained 30 days.
requestBody:
required: true
content:
multipart/form-data:
schema:
type: object
required:
- raw_file
properties:
raw_file:
type: string
format: binary
description: File to parse.
parse_mode:
type: string
enum:
- basic
- standard
default: standard
description: basic (text-only) or standard (multimodal).
enable_document_hierarchy:
type: boolean
description: Add an inferred table of contents (beta).
enable_split_tables:
type: boolean
description: Split large tables by row, repeating headers.
max_split_table_cells:
type: integer
minimum: 1
description: Cell-count threshold above which tables are split.
page_range:
type: string
description: 0-based pages to parse, e.g. "0,1,2" or "0-2,5".
responses:
'200':
description: Parse job accepted.
content:
application/json:
schema:
$ref: '#/components/schemas/ParseJobResponse'
'422':
$ref: '#/components/responses/ValidationError'
/parse/jobs/{job_id}/status:
get:
operationId: parseStatus
tags:
- Parse
summary: Parse Status
parameters:
- $ref: '#/components/parameters/JobId'
responses:
'200':
description: Parse job status.
content:
application/json:
schema:
$ref: '#/components/schemas/ParseStatusResponse'
'404':
description: Job not found or older than 30 days.
'422':
$ref: '#/components/responses/ValidationError'
/parse/jobs/{job_id}/results:
get:
operationId: parseResult
tags:
- Parse
summary: Parse Result
parameters:
- $ref: '#/components/parameters/JobId'
- name: output_types
in: query
required: false
style: form
explode: true
schema:
type: array
items:
type: string
enum:
- markdown-document
- markdown-per-page
- blocks-per-page
responses:
'200':
description: Parse results.
content:
application/json:
schema:
$ref: '#/components/schemas/ParseResultResponse'
'404':
description: Results not found or older than 30 days.
'422':
$ref: '#/components/responses/ValidationError'
/parse/jobs:
get:
operationId: parseListJobs
tags:
- Parse
summary: Parse List Jobs
description: List parse jobs in the workspace.
responses:
'200':
description: A list of parse jobs.
'422':
$ref: '#/components/responses/ValidationError'
/lmunit:
post:
operationId: lmunit
tags:
- LMUnit
summary: LMUnit
description: >-
Evaluate a model response against a natural-language unit test. Total
input is limited to 7000 tokens. Returns a score from 1 (strongly
fails) to 5 (strongly passes).
requestBody:
required: true
content:
application/json:
schema:
$ref: '#/components/schemas/LMUnitRequest'
responses:
'200':
description: Evaluation score.
content:
application/json:
schema:
$ref: '#/components/schemas/LMUnitResponse'
'422':
$ref: '#/components/responses/ValidationError'
/users:
get:
operationId: getUsers
tags:
- Users
summary: Get Users
responses:
'200':
description: A list of workspace users.
'422':
$ref: '#/components/responses/ValidationError'
patch:
operationId: updateUser
tags:
- Users
summary: Update User
requestBody:
required: true
content:
application/json:
schema:
type: object
responses:
'200':
description: User updated.
'422':
$ref: '#/components/responses/ValidationError'
delete:
operationId: removeUser
tags:
- Users
summary: Remove User
parameters:
- name: email
in: query
required: true
schema:
type: string
format: email
responses:
'200':
description: User removed.
'422':
$ref: '#/components/responses/ValidationError'
/users/invite:
post:
operationId: inviteUsers
tags:
- Users
summary: Invite Users
requestBody:
required: true
content:
application/json:
schema:
type: object
properties:
emails:
type: array
items:
type: string
format: email
responses:
'200':
description: Users invited.
'422':
$ref: '#/components/responses/ValidationError'
components:
securitySchemes:
bearerAuth:
type: http
scheme: bearer
description: Contextual AI API key supplied as a Bearer token.
parameters:
AgentId:
name: agent_id
in: path
required: true
schema:
type: string
format: uuid
description: ID of the agent.
DatastoreId:
name: datastore_id
in: path
required: true
schema:
type: string
format: uuid
description: ID of the datastore.
JobId:
name: job_id
in: path
required: true
schema:
type: string
format: uuid
description: ID of the parse job.
responses:
ValidationError:
description: Validation error.
content:
application/json:
schema:
$ref: '#/components/schemas/HTTPValidationError'
schemas:
CreateAgentRequest:
type: object
required:
- name
properties:
name:
type: string
minLength: 3
maxLength: 200
description:
type: string
maxLength: 500
system_prompt:
type: string
maxLength: 8000
no_retrieval_system_prompt:
type: string
maxLength: 8000
multiturn_system_prompt:
type: string
maxLength: 8000
filter_prompt:
type: string
maxLength: 8000
suggested_queries:
type: array
items:
type: string
datastore_ids:
type: array
items:
type: string
format: uuid
template_name:
type: string
CreateAgentResponse:
type: object
properties:
id:
type: string
format: uuid
datastore_ids:
type: array
items:
type: string
format: uuid
EditAgentRequest:
type: object
properties:
name:
type: string
description:
type: string
system_prompt:
type: string
suggested_queries:
type: array
items:
type: string
datastore_ids:
type: array
items:
type: string
format: uuid
Agent:
type: object
properties:
id:
type: string
format: uuid
name:
type: string
description:
type: string
datastore_ids:
type: array
items:
type: string
format: uuid
ListAgentsResponse:
type: object
properties:
agents:
type: array
items:
$ref: '#/components/schemas/Agent'
next_cursor:
type: string
nullable: true
Message:
type: object
required:
- role
- content
properties:
role:
type: string
enum:
- user
- assistant
content:
type: string
QueryRequest:
type: object
required:
- messages
properties:
messages:
type: array
items:
$ref: '#/components/schemas/Message'
description: Messages so far, ending in the latest user message.
stream:
type: boolean
default: false
description: Stream the response as Server-Sent Events.
conversation_id:
type: string
format: uuid
description: Continue an existing conversation.
llm_model_id:
type: string
description: Model ID for a fine-tuned model.
documents_filters:
type: object
description: Metadata filtering criteria for retrieval.
override_configuration:
type: object
description: Temporarily override agent parameters for this query.
QueryResponse:
type: object
properties:
conversation_id:
type: string
format: uuid
message_id:
type: string
format: uuid
message:
$ref: '#/components/schemas/Message'
retrieval_contents:
type: array
items:
type: object
attributions:
type: array
items:
type: object
groundedness_scores:
type: array
items:
type: number
FeedbackRequest:
type: object
required:
- message_id
- feedback
properties:
message_id:
type: string
format: uuid
feedback:
type: string
enum:
- thumbs_up
- thumbs_down
- flagged
- removed
explanation:
type: string
CreateDatastoreRequest:
type: object
required:
- name
properties:
name:
type: string
configuration:
type: object
CreateDatastoreResponse:
type: object
properties:
id:
type: string
format: uuid
Datastore:
type: object
properties:
id:
type: string
format: uuid
name:
type: string
created_at:
type: string
format: date-time
ListDatastoresResponse:
type: object
properties:
datastores:
type: array
items:
$ref: '#/components/schemas/Datastore'
next_cursor:
type: string
nullable: true
IngestDocumentResponse:
type: object
properties:
id:
type: string
format: uuid
description: ID of the document being ingested.
GenerateRequest:
type: object
required:
- model
- messages
- knowledge
properties:
model:
type: string
enum:
- v1
- v2
description: Grounded Language Model (GLM) version.
messages:
type: array
items:
$ref: '#/components/schemas/Message'
description: Conversation messages; the last must be from the user.
knowledge:
type: array
items:
type: string
description: Knowledge sources to ground the response in.
system_prompt:
type: string
avoid_commentary:
type: boolean
default: false
description: Suppress non-grounded conversational text.
temperature:
type: number
minimum: 0
maximum: 1
default: 0
top_p:
type: number
default: 0.9
max_new_tokens:
type: integer
minimum: 1
maximum: 2048
default: 1024
GenerateResponse:
type: object
properties:
response:
type: string
description: The model's answer to the user's final message.
RerankRequest:
type: object
required:
- query
- documents
- model
properties:
query:
type: string
documents:
type: array
items:
type: string
model:
type: string
enum:
- ctxl-rerank-v2-instruct-multilingual
- ctxl-rerank-v2-instruct-multilingual-mini
- ctxl-rerank-v1-instruct
top_n:
type: integer
minimum: 1
instruction:
type: string
description: Natural-language directive steering the ranking.
metadata:
type: array
items:
type: string
description: Per-document metadata, matching the documents array length.
RerankResponse:
type: object
properties:
results:
type: array
items:
type: object
properties:
index:
type: integer
relevance_score:
type: number
format: double
ParseJobResponse:
type: object
properties:
job_id:
type: string
format: uuid
ParseStatusResponse:
type: object
properties:
status:
$ref: '#/components/schemas/ParseJobStatus'
file_name:
type: string
ParseJobStatus:
type: string
enum:
- pending
- processing
- retrying
- completed
- failed
- cancelled
ParseResultResponse:
type: object
properties:
file_name:
type: string
status:
$ref: '#/components/schemas/ParseJobStatus'
markdown_document:
type: string
pages:
type: array
items:
type: object
document_metadata:
type: object
LMUnitRequest:
type: object
required:
- query
- response
- unit_test
properties:
query:
type: string
description: The prompt to which the model responds.
response:
type: string
description: The model response to evaluate.
unit_test:
type: string
description: Natural-language statement to evaluate the response against.
LMUnitResponse:
type: object
properties:
score:
type: number
minimum: 1
maximum: 5
description: Continuous score from 1 (strongly fails) to 5 (strongly passes).
HTTPValidationError:
type: object
properties:
detail:
type: array
items:
$ref: '#/components/schemas/ValidationErrorDetail'
ValidationErrorDetail:
type: object
properties:
loc:
type: array
items:
oneOf:
- type: string
- type: integer
msg:
type: string
type:
type: string