API Documentation

Required headers (JSON endpoints)

Required headers (upload endpoint)

Do not set Content-Type manually for multipart uploads — let the HTTP client set it automatically so the boundary is included.

Request body

POST /api/chat
Content-Type: application/json

{
  "messages": [
    { "role": "user", "content": "What is the capital of France?" }
  ],
  "searchWeb": false
}

Success response — 200 OK

{
  "message": {
    "role": "assistant",
    "content": "The capital of France is Paris."
  },
  "webSearched": false
}

webSearched is true when searchWeb was requested and search results were successfully retrieved and injected into the prompt.

Multi-turn conversation

Pass the full conversation history each time. The backend does not persist history — the client is responsible for maintaining it.

{
  "messages": [
    { "role": "user",      "content": "My name is Alex." },
    { "role": "assistant", "content": "Hello Alex! How can I help?" },
    { "role": "user",      "content": "What is my name?" }
  ]
}

Status codes

Request example

POST /api/chat
Content-Type: application/json

{
  "messages": [
    { "role": "user", "content": "What happened in tech news today?" }
  ],
  "searchWeb": true
}

Response example

{
  "message": {
    "role": "assistant",
    "content": "Based on current web search results: [1] ..."
  },
  "webSearched": true
}

System prompt injected (simplified)

You are a helpful assistant with access to live web search results.
Use the search results below to answer the user's question accurately.
Cite sources when relevant. If the results don't contain enough
information, say so.

--- WEB SEARCH RESULTS ---
[1] Result title
    Snippet text...
    Source: https://example.com

[2] ...
--- END OF SEARCH RESULTS ---

Behaviour when search fails

If DuckDuckGo is unreachable or returns no results, the backend logs the error and falls back to answering without web context (same as searchWeb: false). The response will still have webSearched: false in that case.

cURL example

curl -X POST https://backend.aiassistant.rohitkrishna.online/api/chat \
  -H "Content-Type: application/json" \
  -d '{
    "messages": [{"role":"user","content":"Latest Node.js release?"}],
    "searchWeb": true
  }'

Request

Send a multipart/form-data request with a single file field named document.

POST /api/upload
Content-Type: multipart/form-data; boundary=----boundary

------boundary
Content-Disposition: form-data; name="document"; filename="report.docx"
Content-Type: application/vnd.openxmlformats-officedocument.wordprocessingml.document

<binary file data>
------boundary--

JavaScript fetch example

const form = new FormData();
form.append("document", fileInput.files[0]);

const res = await fetch("https://backend.aiassistant.rohitkrishna.online/api/upload", {
  method: "POST",
  body: form,          // do NOT set Content-Type manually
});
const data = await res.json();
console.log(data.filename, data.characters);

cURL example

curl -X POST https://backend.aiassistant.rohitkrishna.online/api/upload \
  -F "document=@/path/to/report.docx"

Success response — 200 OK

{
  "success": true,
  "filename": "report.docx",
  "characters": 4821,
  "preview": "This report covers Q3 financials..."
}

Supported file types

Status codes

Response — 200 OK

// With knowledge loaded
{
  "loaded": true,
  "characters": 4821
}

// Without knowledge
{
  "loaded": false,
  "characters": 0
}

Response — 200 OK

{
  "success": true,
  "message": "Temporary knowledge cleared"
}

cURL example

curl -X DELETE https://backend.aiassistant.rohitkrishna.online/api/knowledge

Response — 200 OK (Ollama reachable)

{
  "status": "ok",
  "ollama": "connected",
  "model": "llama3.2",
  "modelAvailable": true,
  "availableModels": ["llama3.2:latest", "mistral:7b"]
}

Response — 503 (Ollama unreachable)

{
  "status": "error",
  "ollama": "unreachable",
  "model": "llama3.2"
}

Workflow

1. Call POST /api/upload with the document
2. Use POST /api/chat normally — knowledge is injected automatically
3. Call DELETE /api/knowledge when done to return to general mode

Limitations

• Knowledge is session-scoped — restarting the server clears it
• Only one document can be loaded at a time; uploading replaces the previous one
• Very large documents may exceed the model context window (~8 K tokens for llama3.2)
• Images, charts, and embedded objects in documents are ignored

Python (requests)

import requests

BASE = "https://backend.aiassistant.rohitkrishna.online"

# Basic chat
resp = requests.post(f"{BASE}/api/chat", json={
    "messages": [{"role": "user", "content": "Hello!"}]
})
print(resp.json()["message"]["content"])

# Upload a document
with open("report.docx", "rb") as f:
    resp = requests.post(f"{BASE}/api/upload", files={"document": f})
print(resp.json())

# Chat with document context (same as basic chat — knowledge injected automatically)
resp = requests.post(f"{BASE}/api/chat", json={
    "messages": [{"role": "user", "content": "Summarise the key findings."}]
})
print(resp.json()["message"]["content"])

# Clear knowledge
requests.delete(f"{BASE}/api/knowledge")

JavaScript / Node.js (native fetch)

const BASE = "https://backend.aiassistant.rohitkrishna.online";

// Send a chat message
const chat = await fetch(`${BASE}/api/chat`, {
  method: "POST",
  headers: { "Content-Type": "application/json" },
  body: JSON.stringify({
    messages: [{ role: "user", content: "Hello!" }]
  }),
});
const { message } = await chat.json();
console.log(message.content);

// Upload a file (Node.js 18+ with FormData)
import { readFileSync } from "fs";
const form = new FormData();
form.append("document", new Blob([readFileSync("data.xlsx")]), "data.xlsx");
const upload = await fetch(`${BASE}/api/upload`, { method: "POST", body: form });
console.log(await upload.json());

cURL — full workflow

# 1. Check health
curl https://backend.aiassistant.rohitkrishna.online/api/health

# 2. Chat
curl -X POST https://backend.aiassistant.rohitkrishna.online/api/chat \
  -H "Content-Type: application/json" \
  -d '{"messages":[{"role":"user","content":"What is 2+2?"}]}'

# 3. Upload document
curl -X POST https://backend.aiassistant.rohitkrishna.online/api/upload \
  -F "document=@company_policy.docx"

# 4. Ask about the document
curl -X POST https://backend.aiassistant.rohitkrishna.online/api/chat \
  -H "Content-Type: application/json" \
  -d '{"messages":[{"role":"user","content":"What is the leave policy?"}]}'

# 5. Clear knowledge
curl -X DELETE https://backend.aiassistant.rohitkrishna.online/api/knowledge