Code
Python or TypeScript
Azure runs the model. Doconda builds the file.
Give your Azure OpenAI agent Doconda's tools and it will deliver finished, editable, reviewed Word, Excel, PowerPoint and PDF files.
The model is your deployment
`model` takes the name of your Azure deployment, not the model's name. With Azure AI Foundry, any deployment that supports function calling works.
The variables each client reads
`AzureOpenAI()` reads `AZURE_OPENAI_ENDPOINT`, `AZURE_OPENAI_API_KEY` and `OPENAI_API_VERSION`. In TypeScript, `azure` from `@ai-sdk/azure` reads `AZURE_RESOURCE_NAME` and `AZURE_API_KEY`.
If the model never calls a tool
When the model never uses a tool, the deployment's model doesn't support function calling. Pick one that does and the same code works.
- 01Install the SDKpip install doconda or npm install @doconda/sdk, and your API key.
- 02Give your agent the toolsDoconda's tools, in Azure OpenAI's format.
- 03Your agent delivers the fileYour agent gets the link to the reviewed file.

import json
from doconda import Doconda
from openai import AzureOpenAI
client = AzureOpenAI() # AZURE_OPENAI_ENDPOINT, AZURE_OPENAI_API_KEY, OPENAI_API_VERSION
tools = Doconda().tools()
spec = [{"type": "function", "function": {"name": t.name, "description": t.description,
"parameters": t.input_schema}} for t in tools]
msg = client.chat.completions.create(model="my-deployment", tools=spec, messages=[
{"role": "user", "content": "Write a quarterly sales report in Word"}]).choices[0].message
call = msg.tool_calls[0] # create_document
out = next(t for t in tools if t.name == call.function.name).run(json.loads(call.function.arguments))
out["links"] # {"docx": "https://…", "pdf": "https://…"}
