deepseek-recipe/encoding-decoding-demo

Encoding & Decoding Demo

A web interface for rendering DeepSeek V4.1 prompts, inspecting special tokens, and decoding complete model output into Chat Completions, Responses, and Messages API formats.

Usage

Use the Rust toolchain pinned in the repository and a C/C++ compiler. This example does not require OpenCV, model weights, or an API key. From the repository root:

cargo run -p encoding-decoding-demo --locked

Open http://127.0.0.1:7778. Set DEMO_ADDR to change the listening address.

Use the Encode / Decode buttons to switch between modes; only the selected mode's editor and result are shown. In Encode mode, select an API format, edit its request JSON, and render the request to inspect the model prompt. In Decode mode, edit the complete assistant output and decode it. Model output is rendered with the same special-token highlights as the prompt. Switching modes or formats keeps each request draft and reuses the same model output.

The example output includes <think>…</think> reasoning, answer text, two DSML weather tool calls (Beijing and Shanghai, each with string and numeric arguments), and the end-of-turn marker <｜end▁of▁sentence｜>. Each protocol's sample request declares the weather tool and includes the preceding tool result and follow-up question.

The decoded JSON view shows only response content, reasoning, and tool calls. It omits finish status and server-generated metadata such as response and tool-call IDs, model names, timestamps, token usage, and thinking signatures. Tool inputs and arguments remain intact, including payload fields with those names. The HTTP endpoint still returns the complete protocol response.

The decoder accepts complete assistant content, including the opening <think> when reasoning is present, and an optional leading <｜Assistant｜>. It consumes these frame markers before passing the content to the library's stream parser. Output without an opening <think> is answer content; a leading </think> from a non-thinking assistant prefix is also accepted. Parsing starts outside tool-call markup because the complete output includes its DSML opening block. Tool declarations, JSON output mode, and stop sequences come from the selected protocol request.

Edit the model output, then use the adjacent decode button or Cmd/Ctrl+Enter. The UI uses the default stop finish reason without displaying protocol status. If the pasted output contains <｜end▁of▁sentence｜>, the demo treats its first occurrence as EOS and discards that marker and everything after it.

HTTP endpoints

POST /api/render accepts { "format": "…", "body": { … } } and returns the prompt with highlighted segments.

POST /api/decode accepts the same request plus output and an optional finish_reason (default: stop). The formats are chat_completions, responses, and messages. For example:

curl http://127.0.0.1:7778/api/decode \
  -H 'content-type: application/json' \
  -d '{
    "format": "chat_completions",
    "body": {"messages": [{"role": "user", "content": "Hello"}]},
    "output": "<think>Respond in English.</think>Hello!<｜end▁of▁sentence｜>",
    "finish_reason": "stop"
  }'

The result is { "response": { … }, "segments": [ … ] }. segments preserves the complete supplied model output for highlighting, including its frame markers. response is the complete protocol response, even when the input request sets stream: true. The demo uses the supplied model name or encoding-decoding-demo, a demonstration response ID, and zero token usage because no tokenizer or backend usage data is supplied. Decoding uses the library's stream parser and protocol response accumulators. API callers can supply length or content_filter as the backend finish reason. EOS uses stop instead, and matched request stop sequences take precedence.

This example has no authentication. Keep the default loopback address for local use. It renders prompts and decodes supplied text; it does not execute inference, run tools, or fetch image URLs.