DeepSeek Fill-In-The-Middle (FIM) Completion API
Beta endpoint that performs fill-in-the-middle completions where a prompt and an optional suffix are provided and the model fills the gap. Useful for code completion and inline code generation tasks.
OpenAPI Specification
openapi: 3.1.0
info:
title: DeepSeek FIM Completion
description: The DeepSeek API uses an API format compatible with OpenAI.
version: 1.0.0
servers:
- url: https://api.deepseek.com/beta
paths:
/completions:
post:
summary: DeepSeek Create FIM Completion
description: >-
Generates a fill-in-the-middle completion for a given prompt using
DeepSeek models.
operationId: createFIMCompletion
requestBody:
required: true
content:
application/json:
schema:
type: object
required:
- model
- prompt
properties:
model:
type: string
enum:
- deepseek-chat
description: ID of the model to use.
prompt:
type: string
default: Once upon a time,
description: The prompt to generate completions for.
echo:
type: boolean
nullable: true
description: Echo back the prompt in addition to the completion.
frequency_penalty:
type: number
nullable: true
minimum: -2
maximum: 2
default: 0
description: >
Positive values penalize new tokens based on their
frequency in the text, reducing repetition.
logprobs:
type: integer
nullable: true
maximum: 20
description: >
Number of most likely tokens to return with log
probabilities. The API will always return the log
probability of the sampled token.
max_tokens:
type: integer
nullable: true
description: >-
The maximum number of tokens that can be generated in the
completion.
presence_penalty:
type: number
nullable: true
minimum: -2
maximum: 2
default: 0
description: >
Positive values penalize new tokens that appear in the text,
encouraging discussion of new topics.
stop:
type: object
nullable: true
description: Sequence where the model stops generating tokens.
stream:
type: boolean
nullable: true
description: >
Whether to stream back partial progress. If set, tokens are
sent as server-sent events until completion.
stream_options:
type: object
nullable: true
description: Options for streaming responses.
suffix:
type: string
nullable: true
description: Text that comes after the completion of the inserted text.
temperature:
type: number
nullable: true
maximum: 2
default: 1
description: >-
Controls randomness in generation (higher values = more
random).
top_p:
type: number
nullable: true
maximum: 1
default: 1
description: >
Nucleus sampling parameter. Tokens are selected from the top_p probability mass.
responses:
'200':
description: Successful completion.
content:
application/json:
schema:
type: object
required:
- id
- choices
- created
- model
- object
properties:
id:
type: string
description: Unique identifier for the completion.
choices:
type: array
items:
type: object
description: Array of generated responses.
created:
type: integer
format: int64
description: Unix timestamp of the response creation.
model:
type: string
description: The model used for the completion.
system_fingerprint:
type: string
description: Backend configuration fingerprint (optional).
object:
type: string
enum:
- text_completion
description: Object type of the response.
usage:
type: object
description: Details of token usage.
tags:
- Completions
tags:
- name: Completions