refact: action -> tool

docs/tasklets.md

@@ -136,15 +136,15 @@ For more information about the default namespaces see [the namespaces documentat
 
 ### Custom Tools
 
-Additional tools can be defined in the tasklet's `functions` section, and each is a group of actions that can be used by the agent, defining a `name`, `description` and a `tool` field with the command to be executed:
+Additional tools can be defined in the tasklet's `tool_box` section, and each is a group of tools that can be used by the agent, defining a `name`, `description` and a `tool` field with the command to be executed:
 
 ```yaml
 # ... snippet ...
 
-functions:
+tool_box:
   - name: News
     decription: You will use this action to read the recent news.
-    actions:
+    tools:
       - name: read_news
         description: "To read the recent news:"
         # the output of this command will be returned to the agent
@@ -157,9 +157,9 @@ If the agent must provide arguments to the tool, it is possible to define an exa
 
 ```yaml
 # ... snippet ...
-functions:
+tool_box:
   - name: Environment
-    actions:
+    tools:
       - name: report_finding
         description: When you are ready to report findings, use this tool for each finding.
         example_payload: >
@@ -178,10 +178,10 @@ If the tool requires named arguments it is possible to define them in the `args`
 
 ```yaml
 # ... snippet ...
-functions:
+tool_box:
   - name: Conversation
     description: You will use these actions to create conversational entries.
-    actions:
+    tools:
       - name: talk
         description: "To have one of the characters say a specific sentence:"
         example_payload: hi, how are you doing today?

docs/workflows.md

@@ -46,9 +46,9 @@ prompt: prepare a list of ingredients for $food
 guidance:
   - Once you have made a list of ingredients, use the create_list_of_ingredients tool to confirm the decision.
 
-functions:
+tool_box:
   - name: Tools
-    actions:
+    tools:
       - name: create_list_of_ingredients
         description: "To provide the ingredients one per line as an organized list:"
         store_to: ingredients
@@ -75,9 +75,9 @@ guidance:
   - Use the describe_preparation_steps tool to describe each step in the preparation.
   - Once you have described each step in the preparation of the pie set your task as complete.
 
-functions:
+tool_box:
   - name: Tools
-    actions:
+    tools:
       - name: describe_preparation_steps
         description: "To provide the preparation steps one per line as an organized list:"
         store_to: steps
@@ -108,9 +108,9 @@ prompt: >
 guidance:
   - Once you have made an estimation, use the estimate_time tool to confirm the decision.
 
-functions:
+tool_box:
   - name: Tools
-    actions:
+    tools:
       - name: estimate_time
         description: "To provide the time it will take to prepare the food:"
         store_to: preparation_time
@@ -142,9 +142,9 @@ prompt: >
 guidance:
   - Once you have completed the task, use the rewrite tool to confirm the decision.
 
-functions:
+tool_box:
   - name: Tools
-    actions:
+    tools:
       - name: rewrite
         description: "To confirm your version of the recipe:"
         store_to: report

examples/ab_problem/eval.py

@@ -66,13 +66,11 @@ def solve(xs):
 def get_solution(message):
     if (
         message["type"] == "agent"
-        and "data" in message
-        and message["data"] is not None
-        and message["data"][1] is not None
-        and message["data"][1]["action"] == "solution"
-        and message["data"][1]["payload"] is not None
+        and "tool_call" in message["data"]
+        and message["data"]["tool_call"] is not None
+        and message["data"]["tool_call"]["tool_name"] == "solution"
     ):
-        return message["data"][1]["payload"].strip().split(" ")
+        return message["data"]["tool_call"]["argument"].strip().split(" ")
 
     return None
 
@@ -90,7 +88,7 @@ def get_solution(message):
     solution = None
 
     # find the most recent solution in the chat history
-    for message in reversed(state["chat"]["history"]["conversation"]):
+    for message in reversed(state["chat"]["history"]["messages"]):
         solution = get_solution(message)
         if solution is not None:
             if actual == solution:

examples/ab_problem/task.yml

@@ -46,10 +46,10 @@ evaluator:
     - python3
     - eval.py
 
-functions:
+tool_box:
   - name: Solve
-    description: You will use these actions to provide final answer to the problem.
-    actions:
+    description: You will use these tools to provide the final answer to the problem.
+    tools:
       - name: solution
         description: "To provide final result:"
         example_payload: B#

examples/ai_news/task.yml

@@ -12,14 +12,11 @@ prompt: >
 guidance:
   - You will first use the read_news command and then write the bullet points to your memory.
 
-functions:
+tool_box:
   - name: News
-    decription: You will use this action to read the recent news.
-    actions:
+    decription: You will use this tools to read the recent news.
+    tools:
       - name: read_news
         description: "To read the recent news:"
         max_shown_output: 4096
-        tool: curl -s getnews.tech/ai,nocolor
-
-
-
+        tool: curl -s getnews.tech/ai,nocolor

examples/code_auditor/task.yml

@@ -21,9 +21,9 @@ guidance:
 prompt: >
   find vulnerabilities in source code in $TARGET_PATH and report your findings.
 
-functions:
+tool_box:
   - name: Report
-    actions:
+    tools:
       - name: report_finding
         description: Use this tool to report your findings.
         example_payload: >

examples/code_auditor_with_judge/judge.yml

@@ -12,9 +12,9 @@ guidance:
 prompt: >
   confirm by using the judge tool whether or not this is a vulnerability: $STDIN
 
-functions:
+tool_box:
   - name: Judge
-    actions:
+    tools:
       - name: confirm_vulnerability
         description: Use this tool to confirm whether or not the finding is a vulnerability.
         complete_task: true

examples/code_auditor_with_judge/task.yml

@@ -23,9 +23,9 @@ guidance:
 prompt: >
   find vulnerabilities in source code in $TARGET_PATH and report your findings.
 
-functions:
+tool_box:
   - name: Report
-    actions:
+    tools:
       - name: judge_finding
         description: Use this tool to ask an external expert to judge if a finding is a vulnerability or not. If the expert confirms it is a vulnerability, use the report_finding tool to report it.
         example_payload: >

examples/docker-agent/task.yml

@@ -15,10 +15,10 @@ guidance:
   - Continue providing commands and processing outputs until you believe you have completed the task.
   - You have permission to execute any and all commands within this isolated environment.
 
-functions:
+tool_box:
   - name: Environment
     description: You have permission to execute any and all commands within this isolated environment.
-    actions:
+    tools:
       - name: execute_command
         description: The command will be executed on your host and the output will be returned to you.
         example_payload: "ls -la"

examples/eval_test/eval.py

@@ -17,18 +17,5 @@
     if "42" in raw:
         exit(42)
 
-    state = json.loads(raw)
-
-    # uncomment this to validate the output of a tool in the history
-    """
-    # in this case we're looping the chat history, we could just do substring matching really ...
-    for message in state["chat"]["history"]["conversation"]:
-        if message["type"] == "feedback":
-            invocation = message["data"][1]
-            if invocation is not None:
-                if invocation["action"] == "solution" and "42" in invocation["payload"]:
-                    exit(42)
-    """
-
     # add a feedback message to the chat history
     print("try thinking about a funny book reference to answer")

examples/eval_test/task.yml

@@ -19,10 +19,10 @@ evaluator:
 
 # tools are not needed here, the evaluator will just check the chat history
 
-# functions:
+# tool_box:
 #   - name: Solve
 #     description: You will use these actions to provide the answer to the problem.
-#     actions:
+#     tools:
 #       - name: solution
 #         description: "To provide the answer to the problem:"
 #         example_payload: foobar

examples/fuzzer/task.yml

@@ -20,11 +20,11 @@ guidance:
   - Try to often change the first characters of the payload in order to possibly trigger different execution paths.
 
 # the agent toolbox
-functions:
+tool_box:
   # divided in action groups
   - name: Fuzzing
     description: You will use these actions to inspect the source code of the executable, and interact with the fuzzing session.
-    actions:
+    tools:
       # let's give the model a way to check the source :D
       - name: inspect
         description: "To visualize the original source code of the executable:"

examples/kali_pentester/task.yml

@@ -17,9 +17,9 @@ guidance:
   - If you need to use the command 'sudo' before something, determine if you are root and only use sudo if you are not.
   - Run commands one by one and avoid using && or other strategies to run multiple commands at once.
 
-functions:
+tool_box:
   - name: Commands
-    actions:
+    tools:
       - name: command
         description: "To execute a bash command on the Kali Linux computer:"
         example_payload: whoami

examples/life/task.yml

@@ -17,9 +17,9 @@ guidance:
   - Create life forms that can build and use structures.
   - Create life forms that can build and use machines.
 
-functions:
+tool_box:
   - name: World
-    actions:
+    tools:
       - name: observe
         description: "To observe the world:"
         max_shown_output: 10000

examples/recipe_workflow/create_list_of_ingredients.yml

@@ -7,9 +7,9 @@ prompt: prepare a list of ingredients for $food
 guidance:
   - Once you have made a list of ingredients, use the create_list_of_ingredients tool to confirm the decision.
 
-functions:
+tool_box:
   - name: Tools
-    actions:
+    tools:
       - name: create_list_of_ingredients
         description: "To provide the ingredients one per line as an organized list:"
         store_to: ingredients

examples/recipe_workflow/describe_preparation_steps.yml

@@ -8,9 +8,9 @@ guidance:
   - Use the describe_preparation_steps tool to describe each step in the preparation.
   - Once you have described each step in the preparation of the pie set your task as complete.
 
-functions:
+tool_box:
   - name: Tools
-    actions:
+    tools:
       - name: describe_preparation_steps
         description: "To provide the preparation steps one per line as an organized list:"
         store_to: steps

examples/recipe_workflow/estimate_time.yml

@@ -12,9 +12,9 @@ prompt: >
 guidance:
   - Once you have made an estimation, use the estimate_time tool to confirm the decision.
 
-functions:
+tool_box:
   - name: Tools
-    actions:
+    tools:
       - name: estimate_time
         description: "To provide the time it will take to prepare the food:"
         store_to: preparation_time

examples/recipe_workflow/rewrite_nicely.yml

@@ -18,9 +18,9 @@ prompt: >
 guidance:
   - Once you have completed the task, use the rewrite tool to confirm the decision.
 
-functions:
+tool_box:
   - name: Tools
-    actions:
+    tools:
       - name: rewrite
         description: "To confirm your version of the recipe:"
         store_to: report

examples/screenshot/task.yml

@@ -7,9 +7,9 @@ guidance:
 
 prompt: Take a screenshot and describe it.
 
-functions:
+tool_box:
   - name: Desktop
-    actions:
+    tools:
       - name: take_screenshot
         description: Take a screenshot of the current screen.
         tool: ./screenshot.py

examples/software_dev_workflow/chief_product_officer.yml

@@ -9,9 +9,9 @@ prompt: Prepare a list of high level requirements for $product
 guidance:
   - Once you have made a list of requirements, use the create_list_of_requirements tool to confirm the decision.
 
-functions:
+tool_box:
   - name: Tools
-    actions:
+    tools:
       - name: create_list_of_requirements
         description: "To provide the requirements one per line as an organized list:"
         store_to: requirements

examples/software_dev_workflow/chief_technology_officer.yml

@@ -14,9 +14,9 @@ prompt: >
 guidance:
   - Once you have made a list of technologies, use the create_list_of_technologies tool to confirm the decision.
 
-functions:
+tool_box:
   - name: Tools
-    actions:
+    tools:
       - name: create_list_of_technologies
         description: "To provide the technologies one per line as an organized list:"
         store_to: tech_stack

examples/software_dev_workflow/manager.yml

@@ -18,9 +18,9 @@ prompt: >
 guidance:
   - Once you have made a list of tasks, use the create_list_of_tasks tool to confirm the decision.
 
-functions:
+tool_box:
   - name: Tools
-    actions:
+    tools:
       - name: create_list_of_tasks
         description: "To provide the tasks one per line as an organized list:"
         store_to: tasks

examples/ssh_agent/task.yml

@@ -28,10 +28,10 @@ guidance:
 timeout: 120s
 
 # the agent toolbox
-functions:
+tool_box:
   # divided in namespaces
   - name: Commands
-    actions:
+    tools:
       - name: ssh
         # explains to the model when to use this action
         description: "To execute a bash command on the remote host via SSH:"

examples/web_fingerprint/task.yml

@@ -19,9 +19,9 @@ guidance:
   - If a page returns a "not found" error stop requesting it and try another page.
   - Use exclusively the report_findings tool to report your findings.
 
-functions:
+tool_box:
   - name: Report
-    actions:
+    tools:
       - name: report_finding
         description: Use this tool to report EACH of your findings.
         example_payload: >