Enhance custom tool doc

docs/api/tools/custom-tools.md

@@ -1,7 +1,37 @@
 # Custom Tools
 
-In HybridAGI, each tool is a DSPy module that is optimized in the background when the interpreter call it during optimization. 
-To add a new tool to HybridAGI, you have to create an object that inherit from `BaseTool`:
+In HybridAGI, each tool is a module that is optimized in the background when the interpreter calls it during optimization. Each tool is a DSPy module that is optimized in the background when the interpreter calls it during optimization.
+
+To add a new tool to HybridAGI, there are two ways to do it. The first way is easier, while the other gives you more control but involves more code.
+
+Let's see the easiest way first:
+
+```python
+
+def greet(**kwargs):
+    """
+    This function greets the person passed in as a parameter
+    """
+    name = kwargs["name"] # The function inputs need to use python **kwargs
+    # Here goes your tool logic
+    message = f"Hello, {name}. Nice to meet you!"
+    print(message)
+    # You have then to returns the output of your function as a dict
+    output = {}
+    output["message"] = message
+    return output
+
+my_custom_tool = Tool(
+    name = "Greet", # The name of the tool
+    signature = "name -> message", # The function signature (multi input/output are possible like in DSPy signature)
+    instructions = "Greet a person", # The tool instruction
+    func = greet, # And the callable function
+    lm = my_lm # You can optionally use another LM to infer the inputs of the tool
+)
+
+```
+
+The other way to create a custom tool gives you better control, for that you have to create an object that inherit from `BaseTool`:
 
 ```python
 import abc
@@ -31,7 +61,8 @@ To implement your own tool, here is a toy example:
 ```python
 
 import dspy
-from hybridagi.tools.base import BaseTool
+from typing import Optional
+from hybridagi.tools import BaseTool
 from hybridagi.output_parsers.prediction import PredictionOutputParser
 
 def greet(name):
@@ -53,7 +84,10 @@ class GreetSignature(dspy.Signature):
 
 class GreetTool(BaseTool):
 
-    def __init__(self):
+    def __init__(
+            self,
+            lm = Optional[dspy.LM] = None,
+        ):
         super().__init__(name = "Greet")
         self.predict = dspy.Predict(GreetSignature)
         self.prediction_parser = PredictionOutputParser()
@@ -69,12 +103,15 @@ class GreetTool(BaseTool):
         """Method to perform DSPy forward prediction"""
         if not disable_inference:
             # Perform a DSPy prediction
-            pred = self.predict(
-                objective = objective,
-                context = context,
-                purpose = purpose,
-                prompt = prompt,
-            )
+            with dspy.context(lm=self.lm if self.lm is not None else dspy.settings.lm):
+                # This allow the tool to use a different LM than the one you configured in DSPy
+                # Allowing you to choose the best LM to infer the tool inputs
+                pred = self.predict(
+                    objective = objective,
+                    context = context,
+                    purpose = purpose,
+                    prompt = prompt,
+                )
             # Some parsing to clean up the prediction
             pred.name = self.prediction_parser.parse(pred.name, prefix = "Name:", stop = ["\n"])
             # Then we can call our python function
@@ -92,4 +129,6 @@ class GreetTool(BaseTool):
                 answer = prompt
                 message = message
             )
+
+my_custom_tool = GreetTool()
 ```

static/robot.txt

@@ -0,0 +1,2 @@
+User-Agent: GPTBot
+Disallow: /