summaryrefslogtreecommitdiff
path: root/README.md
diff options
context:
space:
mode:
authorYuval Adam <_@yuv.al>2025-03-06 14:34:00 +0100
committerYuval Adam <_@yuv.al>2025-03-06 14:34:00 +0100
commit6b82841a7042fcd08ec328a0b52cff4e1c9479c9 (patch)
tree8c97a02e7ca6f2ba66119f32e4899131aae0d25b /README.md
parent86be7ada89bf2a4a29d1c6a6ffc60d790e0d5e7a (diff)
Implement core do() step
Diffstat (limited to 'README.md')
-rw-r--r--README.md72
1 files changed, 72 insertions, 0 deletions
diff --git a/README.md b/README.md
index 04ef760..754ea63 100644
--- a/README.md
+++ b/README.md
@@ -31,6 +31,16 @@ uvx strangeloop process FILE_PATH [--output OUTPUT_PATH]
# Ask Claude Sonnet 3.7 a question
uvx strangeloop ask "What is recursive self-improvement in AI?"
+# Use the AI agent loop to fulfill requests
+uvx strangeloop do "find the current weather in New York"
+uvx strangeloop do "create a function to generate secure passwords"
+
+# Manage capabilities
+uvx strangeloop capability add "generate a secure random password"
+uvx strangeloop capability list
+uvx strangeloop capability show generate_secure_password
+uvx strangeloop capability run generate_secure_password 16 --include-special-chars
+
# Configuration management
uvx strangeloop config set anthropic_api_key "your-api-key"
uvx strangeloop config get anthropic_api_key
@@ -49,6 +59,68 @@ strangeloop info
strangeloop ask "What is the meaning of life?" --max-tokens 2048 --temperature 0.8
```
+## AI Agent Loop
+
+The core of Strangeloop is the AI agent loop, which allows you to make requests in natural language and have the system determine the best way to fulfill them:
+
+```bash
+strangeloop do "find information about the weather in Paris"
+```
+
+The system will:
+1. Analyze your request
+2. Check available capabilities
+3. Either:
+ - Use an existing capability
+ - Create a new capability
+ - Provide a direct response
+
+You can control the execution with options:
+```bash
+# Don't automatically execute the suggested action
+strangeloop do "analyze this log file" --no-auto-execute
+
+# Adjust the temperature for more creative responses
+strangeloop do "write a poem about AI" --temperature 0.9
+```
+
+## Capabilities Management
+
+Strangeloop allows you to create, manage, and execute capabilities - Python functions that can be dynamically added to the system:
+
+### Adding Capabilities
+
+```bash
+# Add a new capability
+strangeloop capability add "generate a secure random password"
+
+# Control the generation parameters
+strangeloop capability add "fetch current weather for a location" --temperature 0.5 --max-tokens 4096
+```
+
+### Listing and Viewing Capabilities
+
+```bash
+# List all available capabilities
+strangeloop capability list
+
+# Get detailed information about capabilities
+strangeloop capability list --verbose
+
+# Show details of a specific capability
+strangeloop capability show generate_secure_password
+```
+
+### Running Capabilities
+
+```bash
+# Run a capability with arguments
+strangeloop capability run generate_secure_password 16 true
+
+# Parse arguments as JSON
+strangeloop capability run fetch_weather "New York" --json
+```
+
## API Keys and Configuration
To use the Claude Sonnet 3.7 integration, you need to provide your Anthropic API key in one of these ways (in order of precedence):