How to Chat with All Your PDFs Using the PDF.ai API

What chat-all is for

• You sent a message with your X-API-Key, and no docId

• The response includes a content string with the answer

• The response includes references, and each one names the docId, pageNumber, line range, and url it came from

Practical rule: chat-all needs at least one PDF uploaded through the v1 upload endpoint. It searches what you uploaded, not the open web.

The request at a glance

How to call chat-all

1. Have an API key

2. Upload your PDFs with tags

• By URL: POST https://pdf.ai/api/v1/upload/url with url and tags (300-second timeout)

• By file: POST https://pdf.ai/api/v1/upload/file with file and tags (300-second timeout, 4.5 MB file limit)

import requests

headers = {"X-API-Key": "YOUR_API_KEY"}

response = requests.post(
    "https://pdf.ai/api/v1/upload/url",
    headers=headers,
    json={
        "url": "https://example.com/handbook.pdf",
        "tags": ["HR"]
    }
)
print(response.json())  # {"docId": "..."}

3. Fix tags you forgot

import requests

response = requests.post(
    "https://pdf.ai/api/v1/tags",
    headers={"X-API-Key": "YOUR_API_KEY"},
    json={"docId": "YOUR_DOC_ID", "tags": ["HR", "Policies"]}
)
print(response.json())  # {"message": "Successfully updated tags!"}

4. Example: ask across one tagged set

import requests

url = "https://pdf.ai/api/v1/chat-all"
headers = {
    "X-API-Key": "YOUR_API_KEY",
    "Content-Type": "application/json"
}

data = {
    "message": "How many vacation days do new employees get?",
    "tag": "HR",
    "topMatches": 5
}

response = requests.post(url, headers=headers, json=data)
result = response.json()
print(result["content"])
for ref in result["references"]:
    print(ref["docId"], ref["pageNumber"], ref["fromLine"], ref["toLine"])

curl -X POST https://pdf.ai/api/v1/chat-all \
  -H "X-API-Key: YOUR_API_KEY" \
  -H "Content-Type: application/json" \
  -d '{"message": "How many vacation days do new employees get?", "tag": "HR", "topMatches": 5}'

5. Read the references, not only the answer

Practical rule: if the answer cites documents you did not expect, narrow the call with tag before you change the prompt. A smaller, cleaner set usually beats a cleverer sys_prompt.

6. Handle the documented errors

• 200 — content plus references

• 401 — {"error": "Invalid API key"}: fix the key or header before you touch the request body

• 400 — e.g. {"error": "No API key present"}: the API key or message is missing

Saving chat-all history

Credits and limits before you scale

When chat-all is the wrong tool

• You already know which file holds the answer. Use Chat with one docId (How to Use the PDF.ai Chat API). It skips the library search.

• You want one summary per document. Use Summarize (How to Use the PDF.ai Summarize API).

• You work on v2 parsed documents. Ask (POST https://pdf.ai/api/v2/ask) takes a prompt and one or more parsed docIds (How to Use the PDF.ai Ask API). Do not assume a v1 upload docId works there without checking.

• You need named fields as JSON, not prose. Use POST /v2/extract with a schema (How to Use the PDF.ai Extract API).

• You wanted a generic "pdf api" hub. Stay on https://pdf.ai/developer. This page is the chat-all how-to, not a replacement hub.

A short action plan

1. Open https://pdf.ai/developer, create a key, store it once.

2. Upload your PDFs through v1 with tags (OCR on for scans). Fix missing tags with /v1/tags.

3. POST https://pdf.ai/api/v1/chat-all with message, plus tag and topMatches to keep the search tight.

4. Show references next to every answer, spot-check a few against the source pages, then scale.