Examples / Agents / Multi Model
Multi Model
- Has Specs
- Requires API Keys
Demonstrates using multiple models in a single workflow - a researcher using GPT-4o and a summarizer using GPT-4o-mini. This example shows: - Defining multiple agents with different models - Sequential agent execution (research then summarize) - Template strings in initial messages (`{input.topic}`, `{research}`) - Passing results between agents - Turn-limited loops for each agent phase - Using the done tool to extract agent outputs
Source Code
-- Multi-Model Workflow Example
-- Demonstrates using multiple OpenAI models in one procedure
local done = require("tactus.tools.done")
-- Agents (defined at top level - reusable across procedures)
researcher = Agent {
model = "openai/gpt-4o",
system_prompt = [[You are a researcher. Provide brief research findings (2-3 paragraphs maximum).
IMPORTANT: You MUST call the 'done' tool when finished, passing your research as the 'reason' argument.
]],
initial_message = "Please research this topic and call done when finished: {input.topic}",
tools = {done},
}
summarizer = Agent {
model = "openai/gpt-4o-mini",
system_prompt = [[You are a summarizer. Create a brief 1-2 paragraph summary of the provided text.
IMPORTANT: You MUST call the 'done' tool when finished, passing your summary as the 'reason' argument.
]],
initial_message = "Please summarize the following research and call done when finished:\n\n{research}",
tools = {done},
}
-- Procedure with input defined inline
Procedure {
input = {
topic = field.string{default = "artificial intelligence"},
},
function(input)
-- Research phase with GPT-4o
Log.info("Starting research with GPT-4o...")
local max_turns = 3
local turn_count = 0
local result
repeat
result = researcher()
turn_count = turn_count + 1
until done.called() or turn_count >= max_turns
local research
if done.called() then
research = done.last_result() or "Task completed"
else
if result and result.output ~= nil then
research = tostring(result.output)
else
research = "Research not completed"
end
Log.warn("Researcher did not call done within max turns")
end
state.research = research
-- Reset done tool for next agent
done.reset()
-- Summarization phase with GPT-4o-mini
Log.info("Creating summary with GPT-4o-mini...")
turn_count = 0
repeat
result = summarizer()
turn_count = turn_count + 1
until done.called() or turn_count >= max_turns
local summary
if done.called() then
summary = done.last_result() or "Task completed"
else
if result and result.output ~= nil then
summary = tostring(result.output)
else
summary = "Summary not completed"
end
Log.warn("Summarizer did not call done within max turns")
end
return {
research = research,
summary = summary,
models_used = {"openai/gpt-4o", "openai/gpt-4o-mini"}
}
end
}
-- BDD Specifications
Specification([[
Feature: Multi-Model Workflow
Demonstrate using multiple OpenAI models in one procedure
Scenario: Research and summarization workflow
Given the procedure has started
And the agent "researcher" responds with "I have researched the topic and found key insights about artificial intelligence."
And the agent "researcher" calls tool "done" with args {"reason": "Research findings on artificial intelligence"}
And the agent "summarizer" responds with "Here is a concise summary of the research findings."
And the agent "summarizer" calls tool "done" with args {"reason": "Summary of AI research"}
When the procedure runs
Then the done tool should be called at least 1 time
And the procedure should complete successfully
]])
Quick Start
Run the example:
tactus run 02-agents/01-multi-model.tac
Test with mocks:
tactus test 02-agents/01-multi-model.tac --mock
Note. This example requires API keys. Set your
OPENAI_API_KEYenvironment variable before running.
Explore more examples
Learn Tactus through practical, runnable examples organized by topic.