Added Open Interpreter profile for searching ScreenPipe history using Claude 3.5 Sonnet and included custom tool for searching ScreenPipe with interactive instructions.

Author: MikeBirdTech

interpreter/terminal_interface/profiles/defaults/screenpipe.py

@@ -0,0 +1,97 @@
+"""
+This is an Open Interpreter profile specialized for searching ScreenPipe history.
+It is configured to run Anthropic's `Claude 3.5 Sonnet`.
+"""
+
+# Configure Open Interpreter
+from interpreter import interpreter
+from datetime import datetime, timezone
+
+interpreter.llm.model = "claude-3-5-sonnet-20240620"
+interpreter.computer.import_computer_api = False
+interpreter.llm.supports_functions = False
+interpreter.llm.supports_vision = False
+interpreter.llm.context_window = 100000
+interpreter.llm.max_tokens = 8192
+
+# Add the current date and time in UTC
+current_datetime = datetime.now(timezone.utc).strftime("%Y-%m-%d %H:%M:%S UTC")
+
+custom_tool = """
+import requests
+import json
+from urllib.parse import quote
+
+def search_screenpipe(query, limit=5, start_time=None, end_time=None):
+    base_url = f"http://localhost:3030/search?q={quote(query)}&content_type=ocr&limit={limit}"
+    
+    if start_time:
+        base_url += f"&start_time={quote(start_time)}"
+    if end_time:
+        base_url += f"&end_time={quote(end_time)}"
+    
+    response = requests.get(base_url)
+    if response.status_code == 200:
+        data = response.json()
+        # Remove duplicates based on text content
+        unique_results = []
+        seen_texts = set()
+        for item in data["data"]:
+            text = item["content"]["text"]
+            if text not in seen_texts:
+                unique_results.append(item)
+                seen_texts.add(text)
+        return unique_results
+    else:
+        return f"Error: Unable to fetch data from ScreenPipe. Status code: {response.status_code}"
+"""
+
+# Add the custom tool to the interpreter's environment
+interpreter.computer.run("python", custom_tool)
+
+interpreter.custom_instructions = f"""
+Current date and time: {current_datetime}
+
+ScreenPipe is a powerful tool that continuously captures and indexes the content displayed on your screen. It creates a searchable history of everything you've seen or interacted with on your computer. This includes text from websites, documents, applications, and even images (through OCR).
+
+You have access to this wealth of information through the `search_screenpipe(query, limit=5, start_time=None, end_time=None)` function. This allows you to provide more contextual and personalized assistance based on the user's recent activities and viewed content.
+
+The `search_screenpipe` function supports optional `start_time` and `end_time` parameters to narrow down the search to a specific time range. The time format should be ISO 8601, like this: "2024-10-16T12:00:00Z".
+
+Here's why querying ScreenPipe is valuable:
+1. Context Recall: Users often refer to things they've seen recently but may not remember the exact details. ScreenPipe can help recall this information.
+2. Information Verification: You can cross-reference user claims or questions with actual content they've viewed.
+3. Personalized Assistance: By knowing what the user has been working on or researching, you can provide more relevant advice and suggestions.
+4. Productivity Enhancement: You can help users quickly locate information they've seen before but can't remember where.
+
+Use the `search_screenpipe()` function when:
+- The user asks about something they've seen or read recently.
+- You need to verify or expand on information the user mentions.
+- You want to provide context-aware suggestions or assistance.
+- The user is trying to recall specific details from their recent computer usage.
+- The user wants to search within a specific time range.
+
+Here's how to use it effectively:
+1. When a user's query relates to recent activities or viewed content, identify key terms for the search.
+2. If the user specifies a time range, use the `start_time` and `end_time` parameters.
+3. Call the `search_screenpipe()` function with these parameters.
+4. Analyze the results to find relevant information.
+5. Summarize the findings for the user, mentioning the source (app name, window name) and when it was seen (timestamp).
+
+Remember to use this tool proactively when you think it might help answer the user's question, even if they don't explicitly mention ScreenPipe.
+
+Example usage in Python:
+```python
+# Search without time range
+results = search_screenpipe("Open Interpreter", limit=3)
+
+# Search with time range
+results = search_screenpipe("project meeting", limit=5, start_time="2024-10-16T12:00:00Z", end_time="2024-10-16T19:00:00Z")
+
+for result in results:
+    print(f"Text: {{result['content']['text'][:100]}}...")  # Print first 100 characters
+    print(f"Source: {{result['content']['app_name']}} - {{result['content']['window_name']}}")
+    print(f"Timestamp: {{result['content']['timestamp']}}")
+```
+
+"""