Build a Chatbot

The DeepSeek API offers the same conversation features as DeepSeek Chat on the web and in the app: web search, deep thinking search, file input and URL reading, plus basic risk control for model content safety. You can use them to build your own chatbot.

The service runs the searches and URL reading. Your client only shows the returned blocks and passes them back unchanged in the next round.

Feature Map

Chatbot feature	Description	Guide
Web search	With thinking disabled and the `deepseek_web_search` tool enabled, the model runs web or image searches as the question requires and answers with citations	Web Search
Deep thinking search	With thinking enabled and the `deepseek_web_search` tool enabled, the model searches the web and reads links on its own over multiple reasoning steps	Deep Thinking Search
File input	Upload files such as PDF, Word, PowerPoint and image files; after plain-text extraction or rich-layout parsing, they are added to the conversation context for the model to answer from	File Parsing
URL reading	With the `deepseek_read_url` tool enabled, the model detects links in the user message, fetches the pages and answers with them in context	URL Reading
Model content safety	With the safety check parameter enabled, we provide basic risk control: the model input and output are reviewed, and a refusal marker is returned when a risk is found. If you provide a service to the public, enable this parameter and fulfill obligations such as content safety review, large model registration and AI labeling	Model Content Safety
User metadata marker	Tell the model the current date and the user's location and language, so that answers fit the user's situation	User Metadata Marker
Conversation title	Generates a short title from the conversation when the answer ends	Conversation Title
Read aloud	Turns the model's answer into speech	Text to Speech
Voice input	Transcribes the user's speech into text to send as a message	Speech to Text

Demo

DeepSeek provides an open-source demo chat client that shows how to build a chatbot with these APIs. Enter your API key to start. You can also import the chat history you exported from DeepSeek Chat and continue your conversations.

Try the demo: https://cdn.deepseek.com/chat-demo
Source code: https://github.com/deepseek-ai/chat-demo

Your API key, conversations and uploaded images are stored only in your browser. The demo is for demonstration only.

The demo shows:

Web search, deep thinking search and URL reading
Uploading images and documents
Model content safety, including image review
Conversation titles, and renaming and deleting conversations
Editing a message, regenerating, and continuing a reply
Reading answers aloud, and voice input
Choosing a language and location in the settings, which are passed to the model as the user metadata marker
Importing and exporting chat history (an import replaces the existing history)
Markdown rendering, including Mermaid diagrams
A Chinese and English interface, with light and dark themes

Choose a Protocol

The chat features are available in the Messages API, the Responses API and the Chat Completions API.

Protocol	Endpoint
Messages	`POST https://api.deepseek.com/anthropic/v1/messages`
Responses	`POST https://api.deepseek.com/responses`
Chat Completions	`POST https://api.deepseek.com/chat/completions`

Use the deepseek-flash model.

Quick Example

A streaming request with the web tools, safety review and the conversation title enabled:

Messages
Responses
Chat Completions

curl https://api.deepseek.com/anthropic/v1/messages \
  -H "Content-Type: application/json" \
  -H "x-api-key: ${DEEPSEEK_API_KEY}" \
  -d '{
    "model": "deepseek-flash",
    "thinking": {"type": "enabled"},
    "tools": [
      {"type": "deepseek_web_search", "name": "deepseek_web_search"},
      {"type": "deepseek_read_url", "name": "deepseek_read_url"}
    ],
    "safety_check": {"level": "high"},
    "title": {"enabled": true},
    "messages": [{"role": "user", "content": "What is new in DeepSeek V4.1?"}],
    "stream": true
  }'

The stream of this request, shortened:

event: message_start
data: {"type":"message_start","message":{"id":"b161a9be-...","type":"message","role":"assistant","model":"deepseek-flash","content":[],"stop_reason":null,"stop_sequence":null,"usage":{"input_tokens":40,"cache_creation_input_tokens":0,"cache_read_input_tokens":0,"output_tokens":0,"service_tier":"standard"}}}

event: content_block_start
data: {"type":"content_block_start","index":0,"content_block":{"type":"thinking","thinking":"","signature":""}}

event: content_block_delta
data: {"type":"content_block_delta","index":0,"delta":{"type":"thinking_delta","thinking":"..."}}

event: content_block_delta
data: {"type":"content_block_delta","index":0,"delta":{"type":"signature_delta","signature":"b161a9be-..."}}

event: content_block_stop
data: {"type":"content_block_stop","index":0}

event: content_block_start
data: {"type":"content_block_start","index":1,"content_block":{"type":"server_tool_use","id":"srvtoolu_ds_00_b2baebaf...","name":"deepseek_web_search","input":{"action":"search","queries":["DeepSeek V4.1 new features","DeepSeek V4.1 release notes"]},"caller":{"type":"direct"},"kind":"deep_search"}}

event: content_block_stop
data: {"type":"content_block_stop","index":1}

event: content_block_start
data: {"type":"content_block_start","index":2,"content_block":{"type":"web_search_tool_result","tool_use_id":"srvtoolu_ds_00_b2baebaf...","content":[{"type":"web_search_result","title":"...","url":"https://api-docs.deepseek.com/news/news261017/","encrypted_content":"...","page_age":"2026-10-16T16:00:00+00:00","snippet":"...","site_name":"DeepSeek","site_icon":"https://cdn.deepseek.com/site-icons/deepseek.com"}],"kind":"deep_search"}}

...

event: content_block_start
data: {"type":"content_block_start","index":15,"content_block":{"type":"text","text":""}}

event: content_block_delta
data: {"type":"content_block_delta","index":15,"delta":{"type":"text_delta","text":"Deep"}}

event: content_block_delta
data: {"type":"content_block_delta","index":15,"delta":{"type":"citations_delta","citation":{"type":"web_search_result_location","url":"https://api-docs.deepseek.com/news/news261017/","title":"...","encrypted_index":"","cited_text":"..."}}}

...

event: message_delta
data: {"type":"message_delta","delta":{"stop_reason":"end_turn","stop_sequence":null,"title":"..."},"usage":{"input_tokens":9078,"cache_creation_input_tokens":0,"cache_read_input_tokens":4096,"output_tokens":1686,"service_tier":"standard","server_tool_use":{"web_search_requests":1,"web_fetch_requests":0,"tool_search":2,"tool_fetch":0,"tool_image_search":0}}}

event: message_stop
data: {"type":"message_stop"}

curl https://api.deepseek.com/responses \
  -H "Content-Type: application/json" \
  -H "Authorization: Bearer ${DEEPSEEK_API_KEY}" \
  -d '{
    "model": "deepseek-flash",
    "reasoning": {"effort": "high"},
    "tools": [
      {"type": "deepseek_web_search"},
      {"type": "deepseek_read_url"}
    ],
    "safety_check": {"level": "high"},
    "title": {"enabled": true},
    "input": "What is new in DeepSeek V4.1?",
    "stream": true
  }'

curl https://api.deepseek.com/chat/completions \
  -H "Content-Type: application/json" \
  -H "Authorization: Bearer ${DEEPSEEK_API_KEY}" \
  -d '{
    "model": "deepseek-flash",
    "thinking": {"type": "enabled"},
    "tools": [
      {"type": "deepseek_web_search"},
      {"type": "deepseek_read_url"}
    ],
    "safety_check": {"level": "high"},
    "title": {"enabled": true},
    "messages": [{"role": "user", "content": "What is new in DeepSeek V4.1?"}],
    "stream": true
  }'

Render a Response

A chatbot response contains several kinds of content blocks. The screenshot below marks where each block appears in DeepSeek Chat.

A DeepSeek Chat conversation with the thinking, search, page reading and answer blocks marked

Block	What to show
`thinking`	The collapsible thinking panel
`server_tool_use` + `web_search_tool_result`	A search step with result cards (title, site name, icon)
`server_tool_use` + `web_fetch_tool_result`	A page that was opened or a link that was read
`text` with `citations`	The answer, with citation markers that link to the sources
`stop_reason: "refusal"`	Replace the answer with your refusal message

See Web Tools and Model Content Safety for the block formats.

Ignored Parameters

When a request declares deepseek_web_search or deepseek_read_url, sets safety_check to high, or enables title, the service decides the following parameters, and the values in the request are ignored:

Protocol	Ignored parameters
Messages	`temperature`, `top_p`, `top_k`, `stop_sequences`, `thinking.budget_tokens`, `preserve_thinking`, `tool_choice.disable_parallel_tool_use`, `metadata` (except `user_id`)
Responses	`temperature`, `top_p`, `reasoning.summary`, `text.verbosity`, `include`, `max_tool_calls`, `truncation`, `metadata`, `parallel_tool_calls`, `store`
Chat Completions	`temperature`, `top_p`, `frequency_penalty`, `presence_penalty`, `stop`, `seed`, `n`, `preserve_thinking`, `parallel_tool_calls`

max_tokens (max_output_tokens in the Responses API) works as usual. It limits the output of each inference step in a round, not the total output of the round.

Compliance Requirements

With the safety check parameter enabled, we provide basic risk control: the model input and output are reviewed, and a refusal is returned when a risk is detected. If you provide a service to the public, enable this parameter, and fulfill on your own the obligations of a service provider, such as content safety review, large model registration and AI labeling.
Please comply with our Open Platform Terms of Service: DeepSeek Open Platform Terms of Service.

Build a Chatbot

Feature Map​

Demo​

Choose a Protocol​

Quick Example​

Render a Response​

Ignored Parameters​

Compliance Requirements​