VectorShift API reference

New to Corsair? See API access, authentication, and error handling.

Chatbots

create

chatbots.create Create a chatbot over a pipeline, naming the input and output nodes its messages use Risk: write

await corsair.vectorshift.api.chatbots.create({ pipeline: { id: 'pl_7f3a2b' }, name: 'Support bot', description: 'Answers support questions', deployment_options: {}, access_config: {}, twilio_config: {}, slack_config: {}, deployed: true, input: 'question', output: 'answer' });

Input

Name	Type	Required	Description
`pipeline`	`object`	Yes	—
`name`	`string`	Yes	—
`description`	`string`	Yes	—
`deployment_options`	`Record<string, any>`	Yes	—
`access_config`	`Record<string, any>`	Yes	—
`twilio_config`	`Record<string, any>`	Yes	—
`slack_config`	`Record<string, any>`	Yes	—
`deployed`	`boolean`	Yes	—
`input`	`string`	Yes	The pipeline input node the message is fed to.
`output`	`string`	Yes	The pipeline output node the reply is read from.

pipeline full type

{
  id: string,
  version?: latest | {
    major: number,
    minor: number,
    patch: number
  }
}

Output

Name	Type	Required	Description
`status`	`success \| failed \| string`	Yes	Never `failed` on a returned result: a failed call is thrown as VectorshiftRunFailedError.
`id`	`string \| null`	No	—

run

chatbots.run Send a text or audio message to a chatbot and read its reply, continuing a conversation by id Risk: write

await corsair.vectorshift.api.chatbots.run({ id: 'cb_1a2b3c', text: 'When does my order ship?', conversation_id: 'conv_9f8e7d' });

Input

Name	Type	Required	Description
`id`	`string`	Yes	—
`text`	`string`	No	—
`audio`	`string`	No	Base64-encoded audio.
`conversation_id`	`string`	No	Continue an existing conversation. Omit to start one.

Output

Name	Type	Required	Description
`status`	`success \| failed \| string`	Yes	Never `failed` on a returned result: a failed call is thrown as VectorshiftRunFailedError.
`conversation_id`	`string \| null`	No	Pass this back to continue the conversation.
`output_message`	`string \| null`	No	—
`follow_up_questions`	`string[] \| null`	No	—

Knowledge Bases

create

knowledgeBases.create Create a knowledge base, choosing the file parser, chunk size and overlap Risk: write

await corsair.vectorshift.api.knowledgeBases.create({ name: 'Product docs', description: 'Everything from the help centre', file_processing_implementation: 'LlamaParse', chunk_size: 1000, chunk_overlap: 200 });

Input

Name	Type	Required	Description
`name`	`string`	Yes	—
`description`	`string`	No	—
`file_processing_implementation`	`Default \| Textract \| Unstructured \| LlamaParse \| Other`	No	Which parser reads uploaded files.
`chunk_size`	`number`	No	—
`chunk_overlap`	`number`	No	—
`analyze_documents`	`boolean`	No	—
`apify_key`	`string`	No	Only for the Apify-backed implementation.

Output

Name	Type	Required	Description
`status`	`success \| failed \| string`	Yes	Never `failed` on a returned result: a failed call is thrown as VectorshiftRunFailedError.
`id`	`string \| null`	No	The new knowledge base id.

get

knowledgeBases.get Read one knowledge base by id, or by name with an optional owner Risk: read

await corsair.vectorshift.api.knowledgeBases.get({ id: 'kb_4d2f9a' });

Input

Name	Type	Required	Description
`id`	`string`	No	—
`name`	`string`	No	The knowledge base name, as an alternative to the id.
`username`	`string`	No	Narrows a name lookup to a user. Only with `name`.
`org_name`	`string`	No	Narrows a name lookup to an organization. Only with `name`.

Output

Name	Type	Required	Description
`status`	`success \| failed \| string`	Yes	Never `failed` on a returned result: a failed call is thrown as VectorshiftRunFailedError.
`object`	`Record<string, any> \| null`	No	—

query

knowledgeBases.query Ask one or more questions of a knowledge base, with retrieval, reranking and answering switches Risk: read

await corsair.vectorshift.api.knowledgeBases.query({ id: 'kb_4d2f9a', queries: ['How do I reset my password?'], config: { rerank_documents: true } });

Input

Name	Type	Required	Description
`id`	`string`	Yes	The knowledge base id.
`queries`	`string[]`	Yes	One or more questions, answered against the same retrieval.
`context`	`string`	No	Extra context for the answer.
`search_metadata`	`object`	No	—
`config`	`object`	No	Retrieval and answering switches. Every field is optional.

search_metadata full type

{
  filter?: string,
  opensearch_filter?: string,
  top_k?: number,
  group_by_key?: string
}

config full type

{
  rerank_documents?: boolean,
  generate_metadata_filters?: boolean,
  transform_query?: boolean,
  answer_multi_query?: boolean,
  expand_query?: boolean,
  do_advanced_qa?: boolean,
  format_context_for_llm?: boolean,
  generate_ai_doc_summaries?: boolean,
  retrieval_unit?: string,
  score_cutoff?: number,
  retrieval_config?: {
    max_documents?: number,
    data_fusion_method?: string
  },
  reranking_config?: {
    reranking_model?: string,
    api_key?: string,
    num_chunks_to_rerank?: number
  },
  question_answering_config?: {
    qa_model?: string,
    advanced_qa_mode?: string
  },
  hybrid_search_config?: {
    alpha?: number,
    fusion_method?: string
  }
}

Output

Name	Type	Required	Description
`status`	`success \| failed \| string`	Yes	Never `failed` on a returned result: a failed call is thrown as VectorshiftRunFailedError.
`result`	`Record<string, any> \| null`	No	—

Pipelines

bulkRun

pipelines.bulkRun Run one pipeline once per entry in parallel, with separate inputs and a run id each Risk: write

await corsair.vectorshift.api.pipelines.bulkRun({ id: 'pl_7f3a2b', runs: [{ inputs: { question: 'one' } }, { inputs: { question: 'two' } }] });

Input

Name	Type	Required	Description
`id`	`string`	Yes	The pipeline id to run repeatedly.
`runs`	`object[]`	Yes	One entry per run, each with its own inputs. Run in parallel.

runs full type

{
  inputs?: Record<string, any>
}[]

Output

Name	Type	Required	Description
`status`	`success \| failed \| string`	Yes	Never `failed` on a returned result: a failed call is thrown as VectorshiftRunFailedError.
`run_outputs`	`object[] \| null`	No	—

run_outputs full type

{
  run_id?: string | null,
  outputs?: Record<string, any> | null
}[] | null

create

pipelines.create Create a pipeline from a builder graph, passed through unchanged Risk: write

await corsair.vectorshift.api.pipelines.create({ name: 'Support triage', description: 'Routes tickets by intent', config: { nodes: [], edges: [] } });

Input

Name	Type	Required	Description
`name`	`string`	Yes	—
`description`	`string`	No	—
`config`	`Record<string, any>`	Yes	The pipeline graph. Required by the API but its shape is defined by the builder, so it is passed through unchanged.

Output

Name	Type	Required	Description
`status`	`success \| failed \| string`	Yes	Never `failed` on a returned result: a failed call is thrown as VectorshiftRunFailedError.
`id`	`string \| null`	No	—

list

pipelines.list List pipeline ids, optionally including shared pipelines and the full objects Risk: read

await corsair.vectorshift.api.pipelines.list({ include_shared: true, verbose: true });

Input

Name	Type	Required	Description
`include_shared`	`boolean`	No	Also list pipelines shared with you.
`verbose`	`boolean`	No	Also return the full pipeline objects, not only their ids.

Output

Name	Type	Required	Description
`status`	`success \| failed \| string`	Yes	Never `failed` on a returned result: a failed call is thrown as VectorshiftRunFailedError.
`object_ids`	`string[] \| null`	No	—
`objects`	`Record<string, any>[] \| null`	No	—

run

pipelines.run Run a pipeline with the given inputs and read back every output node value Risk: write

await corsair.vectorshift.api.pipelines.run({ id: 'pl_7f3a2b', inputs: { question: 'Where is my refund?' } });

Input

Name	Type	Required	Description
`id`	`string`	Yes	The pipeline id.
`inputs`	`Record<string, any>`	No	Input name to value, matching the pipeline input nodes.

Output

Name	Type	Required	Description
`status`	`success \| failed \| string`	Yes	Never `failed` on a returned result: a failed call is thrown as VectorshiftRunFailedError.
`run_id`	`string \| null`	No	—
`outputs`	`Record<string, any> \| null`	No	—

Transformations

create

transformations.create Create a transformation from a Python function and its input and output schemas Risk: write

await corsair.vectorshift.api.transformations.create({ name: 'Order total', description: 'Sums the line items', function_name: 'total', function: 'def total(items):\n    return sum(items)', input_schema: { items: 'list[float]' }, output_schema: { total: 'float' } });

Input

Name	Type	Required	Description
`name`	`string`	Yes	—
`description`	`string`	Yes	—
`function_name`	`string`	Yes	The Python function to call.
`function`	`string`	Yes	The function source.
`input_schema`	`Record<string, string>`	Yes	Argument name to type, for the generated signature.
`output_schema`	`Record<string, string>`	Yes	—

Output

Name	Type	Required	Description
`status`	`success \| failed \| string`	Yes	Never `failed` on a returned result: a failed call is thrown as VectorshiftRunFailedError.
`id`	`string \| null`	No	—

run

transformations.run Run a transformation in VectorShift’s sandbox with the given arguments Risk: write

await corsair.vectorshift.api.transformations.run({ id: 'tr_5e6f7a', inputs: { items: [19.5, 22.5] } });

Input

Name	Type	Required	Description
`id`	`string`	Yes	—
`inputs`	`Record<string, string \| number \| boolean \| any[] \| Record<string, any>>`	Yes	Argument name to value, matching the transformation input schema.

Output

Name	Type	Required	Description
`status`	`success \| failed \| string`	Yes	Never `failed` on a returned result: a failed call is thrown as VectorshiftRunFailedError.
`result`	`Record<string, any> \| null`	No	—

Overview

Plugins

VectorShift API reference

Chatbots

create

run

Knowledge Bases

create

get

query

Pipelines

bulkRun

create

list

run

Transformations

create

run

​Chatbots

​create

​run

​Knowledge Bases

​create

​get

​query

​Pipelines

​bulkRun

​create

​list

​run

​Transformations

​create

​run

Chatbots

create

run

Knowledge Bases

create

get

query

Pipelines

bulkRun

create

list

run

Transformations

create

run