feat: add chat with pdf integration

integrations/chat-with-pdf/README.md

@@ -0,0 +1,51 @@
+# Chat With PDF Agent
+
+## Description
+This code defines a Python module for an AI agent that interacts with a PDF using the pdf.ai API. The agent can receive a question and a URL of a PDF, upload the PDF if not already uploaded, and then retrieve an answer from the PDF.
+
+## Authentication
+
+To use the PDF.ai Summary API, you need to generate an API key. Follow the steps below to obtain your API key:
+
+1. Visit the PDF.ai API Portal: [API Portal Link](https://pdf.ai/api-portal)
+2. Sign in to your account or create a new one.
+3. Navigate to the "API Keys" section in your account dashboard.
+4. Click on "Generate New API Key."
+5. Provide a name for your API key to easily identify its purpose.
+6. Click "Generate Key" to create the API key.
+7. Copy the generated API key and securely store it.
+
+# Agent Secrets on Agentverse
+
+1. Go to the Agentverse platform.
+2. Navigate to the Agent Secrets section.
+3. Create an agent and copy the code in it
+4. Add a new secret with the key `API_KEY` and the value as your API KEY.
+
+# Steps to Enroll an Agent as a Service on Agentverse
+
+You can integrate into DeltaV your Agents created on your local computer, IoT devices, in the VMs, or agents created on Agentverse. The steps are the same.
+
+Once your agents are run, the agent protocol manifests are uploaded to the Almanac contract in the form of protocol digests. After uploading the manifests, we take the agent addresses and enroll the agents as a service under the "Services" tab in Agentverse.
+
+## Agent Validation on Agentverse Explorer
+*Note: You can validate the procedure by searching for your agent's address on Agent Explorer, checking if the protocols have been uploaded successfully. If not, you need to wait for some time (1-2 minutes) until the protocols are uploaded successfully.*
+
+## Create a Service Group
+
+1. Start by creating a new service group on Agentverse.
+2. Set up the service group as PRIVATE (you will only be able to see your own agents).
+   - If you set up your service group as Public, anyone will be able to see your agents.
+
+**Service group has been created.**
+
+## Create a Service
+
+1. To register the agents as a service, input a concise title and description for the agent service.
+2. Choose the service group for the agent service that you've created previously.
+3. Fill in the agent address in the Agent field.
+4. Set the task type to Task.
+
+![Image](./image.png)
+
+Now, your agents are enrolled as a service in Agentverse. You can manage and monitor them under the "Services" tab. Ensure that you follow the agent validation steps on Agent Explorer to confirm successful enrollment.

integrations/chat-with-pdf/agent.py

@@ -0,0 +1,85 @@
+from ai_engine import UAgentResponse, UAgentResponseType
+
+# Define the Chat With PDF Request model
+class ChatWithPDFRequest(Model):
+    url: str
+    question: str
+
+# Define the protocol for Chatting With PDF
+chat_with_pdf_protocol = Protocol("Chat With PDF")
+
+# Dictionary to store URL and corresponding docId
+url_docId_map = {}
+
+def upload_pdf_if_needed(url, ctx):
+    """Uploads a PDF from a URL if it's not already uploaded, and returns the document ID."""
+    if url in url_docId_map:
+        ctx.logger.info(f"PDF is already uploaded")
+        return url_docId_map[url]
+
+    endpoint = "https://pdf.ai/api/v1/upload/url"
+    headers = {"X-API-Key": API_KEY}
+    payload = {"url": url, "isPrivate": False, "ocr": False}
+
+    try:
+        response = requests.post(endpoint, json=payload, headers=headers)
+        response.raise_for_status()
+        docId = response.json().get("docId")
+        url_docId_map[url] = docId  # Store the docId for future reference
+        return docId
+    except Exception as e:
+        ctx.logger.info(f"Error during PDF upload: {e}")
+        return None
+
+def chat_with_pdf(doc_id, question, ctx):
+    """Sends a question to the PDF and returns the answer."""
+    endpoint = "https://pdf.ai/api/v1/chat"
+    headers = {"X-API-Key": API_KEY}
+    payload = {"docId": doc_id, "message": question, "save_chat": False, "language": "english", "use_gpt4": False}
+
+    try:
+        response = requests.post(endpoint, json=payload, headers=headers)
+        ctx.logger.info(f"Chat with PDF Response Status: {response.status_code}")
+        ctx.logger.info(f"Chat with PDF Response Body: {response.text}")
+        response.raise_for_status()
+        content = response.json().get("content", None)
+        return content
+    except Exception as e:
+        ctx.logger.info(f"Error during chat with PDF: {e}")
+        return None
+
+@chat_with_pdf_protocol.on_message(model=ChatWithPDFRequest, replies=UAgentResponse)
+async def on_message(ctx: Context, sender: str, msg: ChatWithPDFRequest):
+    ctx.logger.info(f"Received chat with PDF request from {sender}.")
+
+    try:
+        ctx.logger.info("Checking if PDF needs uploading")
+        doc_id = upload_pdf_if_needed(msg.url, ctx)
+        if not doc_id:
+            raise Exception("Failed to upload or retrieve PDF.")
+
+        ctx.logger.info("Chatting with the PDF")
+        answer = chat_with_pdf(doc_id, msg.question, ctx)
+        if not answer:
+            raise Exception("Failed to chat with PDF.")
+
+        await ctx.send(
+            sender,
+            UAgentResponse(
+                message=answer,
+                type=UAgentResponseType.FINAL
+            )
+        )
+
+    except Exception as exc:
+        ctx.logger.error(f"An error occurred: {exc}")
+        await ctx.send(
+            sender,
+            UAgentResponse(
+                message=f"Error: {exc}",
+                type=UAgentResponseType.ERROR
+            )
+        )
+
+# Include the Chat With PDF protocol in your agent
+agent.include(chat_with_pdf_protocol)

integrations/chat-with-pdf/project.json

@@ -0,0 +1,6 @@
+{
+    "title": "Chat with PDF",
+    "description": "This code defines for an AI agent that interacts with a PDF using the pdf.ai API. The agent can receive a question and a URL of a PDF, upload the PDF if not already uploaded, and then retrieve an answer from the PDF",
+    "categories": ["Text Summarization"],
+    "deltav": true
+}