Skip to main content

​
POST /api/chat

Streams responses from the AI assistant powered by Ollama. Supports automatic context injection based on question content.

​
Request Body

​
model
string
required
The Ollama model to use (e.g., “llama2”, “mistral”, “medllama2”)
​
prompt
string
required
The user’s question or message to the AI
​
useContext
boolean
default:"automatic"
Whether to inject medical context. If not specified, automatically determined based on keywords in the prompt
​
instruction
string
Custom system instruction for the AI. Defaults to: “Responda com base EXCLUSIVA no contexto fornecido…”
​
maxItems
integer
default:"500"
Maximum number of context items to include (when context is used)
​
maxChars
integer
default:"12000"
Maximum characters of context to include (when context is used)

​
Response

The response is streamed using Server-Sent Events (SSE) format with Content-Type: application/octet-stream. Each line contains a JSON object with the AI’s response chunk:
​
response
string
Partial text response from the AI model
​
done
boolean
Indicates if the response is complete

​
Example Request

cURL
curl -X POST http://localhost:8080/api/chat \
  -H "Content-Type: application/json" \
  -d '{
    "model": "llama2",
    "prompt": "O que é diabetes?"
  }' \
  --no-buffer
JavaScript
const response = await fetch('http://localhost:8080/api/chat', {
  method: 'POST',
  headers: { 'Content-Type': 'application/json' },
  body: JSON.stringify({
    model: 'llama2',
    prompt: 'O que é diabetes?'
  })
});

const reader = response.body.getReader();
const decoder = new TextDecoder();

while (true) {
  const { done, value } = await reader.read();
  if (done) break;
  
  const chunk = decoder.decode(value);
  const lines = chunk.split('\n').filter(line => line.trim());
  
  for (const line of lines) {
    const data = JSON.parse(line);
    process.stdout.write(data.response || '');
  }
}

​
Example Response Stream

{"model":"llama2","created_at":"2024-03-15T10:30:00Z","response":"Diabetes","done":false}
{"model":"llama2","created_at":"2024-03-15T10:30:00Z","response":" é uma","done":false}
{"model":"llama2","created_at":"2024-03-15T10:30:00Z","response":" doença crônica","done":false}
...
{"model":"llama2","created_at":"2024-03-15T10:30:01Z","response":".","done":true}

​
Automatic Context Detection

The system automatically detects if the question would benefit from medical context based on keywords: Context Keywords:
  • CID-related: “cid”, “código”, “doença”, “diagnóstico”, “classificação”
  • News-related: “notícia”, “news”, “governo”, “saúde pública”, “ministério”
  • Medical skills: “habilidade”, “médica”, “especialidade”, “cfm”, “resolução”
  • Vigilance: “vigilância”, “guia”, “termo”, “definição”
  • General medical: “sintomas”, “tratamento”, “prevenção”, “epidemiologia”
When context keywords are detected, the system:
  1. Builds relevant context from the medical knowledge base
  2. Injects it into the prompt before sending to Ollama
  3. Instructs the AI to answer based on the provided context

​
Implementation Details

From OllamaChatController.java:27-106: 
@PostMapping(value = "/chat", produces = MediaType.APPLICATION_OCTET_STREAM_VALUE)
public ResponseBodyEmitter streamOllama(@RequestBody ChatRequest request) {
    // Determines if context should be used
    boolean shouldUseContext = Boolean.TRUE.equals(request.getUseContext()) || 
                              shouldUseContextAutomatically(request.getPrompt());
    
    if (shouldUseContext) {
        // Builds context and injects into prompt
        String contexto = contextService.buildContextFromQuestion(...);
        finalPrompt = instruction + "[CONTEXTO]\n" + contexto + 
                     "[PERGUNTA]\n" + prompt;
    }
    // Streams response from Ollama
}

​
Use Cases

  • Medical consultation assistance
  • Patient education and information
  • Disease and symptom lookup
  • Treatment information
  • Medical news and updates
  • General health questions
Prerequisites:
  • Ollama must be running locally at http://localhost:11434
  • The specified model must be pulled/available in Ollama
  • Install Ollama: https://ollama.ai
For explicitly providing full medical context, use POST /api/chat/with-context instead.

​
Error Responses

Status CodeDescription
200Streaming response in progress
400Invalid request (missing model or prompt)
500Error connecting to Ollama or streaming failed
The streaming response allows for real-time display of the AI’s answer as it’s being generated, providing a better user experience.

Build docs developers (and LLMs) love

Get started for freeTalk to us