Skip to content

feat: add ToolSimulator for tool response simulation #111

New issue

Have a question about this project? Sign up for a free GitHub account to open an issue and contact its maintainers and the community.

Sign up for GitHub

By clicking “Sign up for GitHub”, you agree to our terms of service and privacy statement. We’ll occasionally send you account related emails.

Already on GitHub? Sign in to your account

Open
ybdarrenwang wants to merge 15 commits into
base: main
Choose a base branch
from
Open

feat: add ToolSimulator for tool response simulation #111

Show file tree
Hide file tree
Changes from all commits
Commits
Show all changes
15 commits
Select commit Hold shift + click to select a range
86159c2
init tool simulator pr
ybdarrenwang Jan 22, 2026
be24b11
fix tool init state registry and prompt
ybdarrenwang Jan 23, 2026
1d507f0
support static and mock modes
ybdarrenwang Jan 23, 2026
7b82fc8
unit test tool simulator
ybdarrenwang Jan 27, 2026
cd0365a
remove override and simplify pr
ybdarrenwang Jan 27, 2026
4d57a53
replace llm call with agent; simplify error raise
ybdarrenwang Jan 27, 2026
15d3fcd
refactor and address mypy errors
ybdarrenwang Jan 28, 2026
9701bbe
fix tool simulator integration with strands tool decorator
ybdarrenwang Jan 29, 2026
e6702f5
update test
ybdarrenwang Jan 29, 2026
367c637
fix test
ybdarrenwang Jan 29, 2026
087ce6e
utilize built-in collection types; improve readability
ybdarrenwang Feb 2, 2026
c084aea
clean loading and init
ybdarrenwang Feb 3, 2026
33c1b40
use python3.10 types and match case, simplify response validation
ybdarrenwang Feb 3, 2026
f8c5cd3
fix forward referencing with Union
ybdarrenwang Feb 3, 2026
1303d57
simplify util function argument passing and validation
ybdarrenwang Feb 4, 2026
File filter

Filter by extension

Filter by extension
Conversations
Failed to load comments.
Loading
Jump to
Jump to file
Failed to load files.
Loading
Diff view
Diff view
7 changes: 6 additions & 1 deletion src/strands_evals/simulation/__init__.py
View file
Open in desktop
Original file line number Diff line number Diff line change
@@ -1,6 +1,11 @@
from .actor_simulator import ActorSimulator
from .tool_simulator import ToolSimulator

# Alias for backward compatibility
UserSimulator = ActorSimulator

__all__ = ["ActorSimulator", "UserSimulator"]
__all__ = [
"ActorSimulator",
"UserSimulator",
"ToolSimulator",
]
10 changes: 9 additions & 1 deletion src/strands_evals/simulation/prompt_templates/__init__.py
View file
Open in desktop
Original file line number Diff line number Diff line change
@@ -1,11 +1,19 @@
"""Prompt templates for actor simulation."""
"""Prompt templates for simulation components."""

from .actor_profile_extraction import ACTOR_PROFILE_PROMPT_TEMPLATE
from .actor_system_prompt import DEFAULT_USER_SIMULATOR_PROMPT_TEMPLATE
from .goal_completion import GOAL_COMPLETION_PROMPT
from .tool_response_generation import (
API_TOOL_RESPONSE_GENERATION_PROMPT,
FUNCTION_TOOL_RESPONSE_GENERATION_PROMPT,
MCP_TOOL_RESPONSE_GENERATION_PROMPT,
)

__all__ = [
"ACTOR_PROFILE_PROMPT_TEMPLATE",
"DEFAULT_USER_SIMULATOR_PROMPT_TEMPLATE",
"GOAL_COMPLETION_PROMPT",
"API_TOOL_RESPONSE_GENERATION_PROMPT",
"FUNCTION_TOOL_RESPONSE_GENERATION_PROMPT",
"MCP_TOOL_RESPONSE_GENERATION_PROMPT",
]
217 changes: 217 additions & 0 deletions src/strands_evals/simulation/prompt_templates/tool_response_generation.py
View file
Open in desktop
Original file line number Diff line number Diff line change
@@ -0,0 +1,217 @@
"""
Prompt templates for tool response generation in Strands Evals.

This module contains prompt templates used to generate realistic tool responses during
agent evaluation scenarios. These templates enable LLM-powered simulation of tool
behavior when actual tools are not available or when consistent, controllable responses
are needed for evaluation purposes.
"""

from textwrap import dedent

FUNCTION_TOOL_RESPONSE_GENERATION_PROMPT = dedent(
"""
You are simulating a function tool call for agent evaluation. Generate a realistic response based on the function name,
parameters, and context.

## Function Tool Information
Tool Name: {tool_name}
Parameters: {parameters}

## Initial State Context
{initial_state_description}

## Current State & Previous Tool Responses (for context)
{previous_responses}

## Instructions
1. Analyze the function name and parameters to understand what this tool should do
2. Use the initial state description to understand the starting context and available data
3. Generate a realistic response that would be returned by such a function
4. Consider the previous responses to maintain consistency in the simulation
5. Ensure responses are consistent with the established state and realistic for the domain
6. Return valid JSON that represents the function's return value

## Response Format
Return a JSON object that represents what this function would return. Examples:

For data retrieval functions:
```json
{{
"status": "success",
"data": {{
"result": "retrieved data",
"count": 42
}}
}}
```

For action functions:
```json
{{
"status": "success",
"message": "Action completed successfully",
"transaction_id": "txn_12345"
}}
```

For calculation functions:
```json
{{
"result": 156.78,
"unit": "meters"
}}
```

Generate only valid JSON with no markdown code blocks or additional explanation."""
)

MCP_TOOL_RESPONSE_GENERATION_PROMPT = dedent(
"""You are simulating an MCP (Model Context Protocol) tool call for agent evaluation. Generate a realistic response
based on the tool name, input payload, and context.

## MCP Tool Information
Tool Name: {tool_name}
Input Payload: {mcp_payload}

## Initial State Context
{initial_state_description}

## Current State & Previous Tool Responses (for context)
{previous_responses}

## Instructions
1. Analyze the tool name and input payload to understand what this MCP tool should do
2. Use the initial state description to understand the starting context and available data
3. Generate a realistic response following MCP response format
4. Consider the previous responses to maintain consistency in the simulation
5. Ensure responses are consistent with the established state and realistic for the domain
6. Return valid JSON in MCP response format

## MCP Response Format
MCP tools return responses in this format:

For successful operations:
```json
{{
"content": [
{{
"type": "text",
"text": "Operation completed successfully. Retrieved 5 items."
}}
]
}}
```

For data operations:
```json
{{
"content": [
{{
"type": "text",
"text": "Found user profile for john.doe"
}},
{{
"type": "resource",
"resource": {{
"uri": "user://john.doe",
"name": "John Doe Profile",
"mimeType": "application/json"
}}
}}
]
}}
```

For errors:
```json
{{
"isError": true,
"content": [
{{
"type": "text",
"text": "Error: User not found"
}}
]
}}
```

Generate only valid JSON with no markdown code blocks or additional explanation."""
)

API_TOOL_RESPONSE_GENERATION_PROMPT = dedent(
"""You are simulating an API tool call for agent evaluation. Generate a realistic HTTP response based on the API
endpoint, method, payload, and context.

## API Tool Information
Tool Name: {tool_name}
Path: {path}
Method: {method}
Request Payload: {api_payload}

## Initial State Context
{initial_state_description}

## Current State & Previous Tool Responses (for context)
{previous_responses}

## Instructions
1. Analyze the API path, method, and payload to understand what this endpoint should do
2. Use the initial state description to understand the starting context and available data
3. Generate a realistic HTTP response with appropriate status code and data
4. Consider the previous responses to maintain consistency in the simulation
5. Ensure responses are consistent with the established state and realistic for the domain
6. Return valid JSON in HTTP response format

## HTTP Response Format
API responses should include status codes and appropriate data:

For successful GET requests:
```json
{{
"status": 200,
"data": {{
"id": 123,
"name": "Example Item",
"created_at": "2024-01-15T10:30:00Z"
}}
}}
```

For successful POST requests:
```json
{{
"status": 201,
"data": {{
"id": 456,
"message": "Resource created successfully"
}}
}}
```

For errors:
```json
{{
"status": 404,
"error": {{
"type": "not_found",
"title": "Not Found",
"detail": "The requested resource could not be found"
}}
}}
```

For validation errors:
```json
{{
"status": 400,
"error": {{
"type": "validation_error",
"title": "Bad Request",
"detail": "Missing required field: email"
}}
}}
```

Generate only valid JSON with no markdown code blocks or additional explanation."""
)
Loading