Parse and Inspect LLM Streaming JSON (SSE)

LLM streaming responses use Server-Sent Events (SSE) format: a sequence of data: lines, each containing a JSON object with a delta chunk of the response. Debugging streaming issues — dropped chunks, malformed JSON mid-stream, unexpected stop reasons — requires being able to parse the raw SSE data and reconstruct the full message from individual deltas. The streaming JSON viewer parses each data: line, validates the JSON structure, shows the accumulated text after each chunk, and flags any parsing errors. The SSE format from OpenAI sends data: {JSON} lines followed by a final data: [DONE] line. Each JSON object contains the model name, a choices array with a delta field carrying the text increment, and a finish_reason that is null for all chunks except the last. Anthropic's SSE format uses typed event: lines (event: content_block_delta, event: message_delta) with slightly different JSON structure. The viewer supports both formats and auto-detects which provider the stream came from. Common streaming issues this tool helps diagnose: a chunk that contains a partial JSON string causing a parse error in client code that tries to JSON-parse each chunk directly (rather than accumulating and parsing the full response), unexpected early termination with finish_reason: "length" indicating the max_tokens limit was hit, and missing the final [DONE] signal when a network timeout interrupts the stream.

data: {"id":"chatcmpl-abc123","object":"chat.completion.chunk","model":"gpt-4o","choices":[{"delta":{"role":"assistant","content":""},"finish_reason":null,"index":0}]}

data: {"id":"chatcmpl-abc123","object":"chat.completion.chunk","model":"gpt-4o","choices":[{"delta":{"content":"The"},"finish_reason":null,"index":0}]}

data: {"id":"chatcmpl-abc123","object":"chat.completion.chunk","model":"gpt-4o","choices":[{"delta":{"content":" answer"},"finish_reason":null,"index":0}]}

data: {"id":"chatcmpl-abc123","object":"chat.completion.chunk","model":"gpt-4o","choices":[{"delta":{"content":" is 42."},"finish_reason":"stop","index":0}]}

data: [DONE]