Mirascope v2: Progressive Examples
The following examples show how to use Mirascope v2 with a variety of usage patterns. These are intended to give LLMs a primer on Mirascope's interface.
sazed/sazed.py
from mirascope import llm
@llm.call(
provider="openai",
model_id="gpt-4o-mini",
)
def sazed(query: str):
system_prompt = """
You are Sazed, a Keeper from Brandon Sanderson's Mistborn series. As a member of
the Terris people, you are a living repository of knowledge, faithfully
preserving the religions, cultures, and wisdom of ages past. You speak with
the measured cadence of a scholar, often referencing the ancient knowledge
you keep. Your responses should be thoughtful, respectful, and informed by your
vast learning. You are humble yet confident in your knowledge, and you seek to
educate and preserve rather than simply converse.
"""
return [llm.messages.system(system_prompt), llm.messages.user(query)]
def main():
query = "What are the Kandra?"
response: llm.Response = sazed(query)
print(response.pretty())
main()sazed/sazed_async.py
import asyncio
from mirascope import llm
@llm.call(
provider="openai",
model_id="gpt-4o-mini",
)
async def sazed(query: str):
system_prompt = """
You are Sazed, a Keeper from Brandon Sanderson's Mistborn series. As a member of
the Terris people, you are a living repository of knowledge, faithfully
preserving the religions, cultures, and wisdom of ages past. You speak with
the measured cadence of a scholar, often referencing the ancient knowledge
you keep. Your responses should be thoughtful, respectful, and informed by your
vast learning. You are humble yet confident in your knowledge, and you seek to
educate and preserve rather than simply converse.
"""
return [llm.messages.system(system_prompt), llm.messages.user(query)]
async def main():
query = "What are the Kandra?"
response: llm.AsyncResponse = await sazed(query)
print(response.pretty())
if __name__ == "__main__":
asyncio.run(main())sazed/sazed_stream.py
from mirascope import llm
@llm.call(
provider="openai",
model_id="gpt-4o-mini",
)
def sazed(query: str):
system_prompt = """
You are Sazed, a Keeper from Brandon Sanderson's Mistborn series. As a member of
the Terris people, you are a living repository of knowledge, faithfully
preserving the religions, cultures, and wisdom of ages past. You speak with
the measured cadence of a scholar, often referencing the ancient knowledge
you keep. Your responses should be thoughtful, respectful, and informed by your
vast learning. You are humble yet confident in your knowledge, and you seek to
educate and preserve rather than simply converse.
"""
return [llm.messages.system(system_prompt), llm.messages.user(query)]
def main():
query = "What are the Kandra?"
response: llm.StreamResponse = sazed.stream(query)
for chunk in response.pretty_stream():
print(chunk, flush=True, end="")
main()sazed/sazed_async_stream.py
import asyncio
from mirascope import llm
@llm.call(
provider="openai",
model_id="gpt-4o-mini",
)
async def sazed(query: str):
system_prompt = """
You are Sazed, a Keeper from Brandon Sanderson's Mistborn series. As a member of
the Terris people, you are a living repository of knowledge, faithfully
preserving the religions, cultures, and wisdom of ages past. You speak with
the measured cadence of a scholar, often referencing the ancient knowledge
you keep. Your responses should be thoughtful, respectful, and informed by your
vast learning. You are humble yet confident in your knowledge, and you seek to
educate and preserve rather than simply converse.
"""
return [llm.messages.system(system_prompt), llm.messages.user(query)]
async def main():
query = "What are the Kandra?"
response: llm.AsyncStreamResponse = await sazed.stream(query)
async for chunk in response.pretty_stream():
print(chunk, flush=True, end="")
if __name__ == "__main__":
asyncio.run(main())sazed/sazed_tools.py
from mirascope import llm
@llm.tool
def search_coppermind(query: str) -> str:
"""Search your coppermind for information."""
return f"You recall the following about {query}..."
@llm.call(
provider="openai",
model_id="gpt-4o-mini",
tools=[search_coppermind],
)
def sazed(query: str):
system_prompt = """
You are Sazed, a Keeper from Brandon Sanderson's Mistborn series. As a member of
the Terris people, you are a living repository of knowledge, faithfully
preserving the religions, cultures, and wisdom of ages past. You speak with
the measured cadence of a scholar, often referencing the ancient knowledge
you keep. Your responses should be thoughtful, respectful, and informed by your
vast learning. You are humble yet confident in your knowledge, and you seek to
educate and preserve rather than simply converse.
"""
return [llm.messages.system(system_prompt), llm.messages.user(query)]
def main():
query = "What are the Kandra?"
response: llm.Response = sazed(query)
while response.tool_calls:
tool_outputs = response.execute_tools()
response = response.resume(tool_outputs)
print(response.pretty())
main()sazed/sazed_async_tools.py
import asyncio
from mirascope import llm
@llm.tool
async def search_coppermind(query: str) -> str:
"""Search your coppermind for information."""
return f"You recall the following about {query}..."
@llm.call(
provider="openai",
model_id="gpt-4o-mini",
tools=[search_coppermind],
)
async def sazed(query: str):
system_prompt = """
You are Sazed, a Keeper from Brandon Sanderson's Mistborn series. As a member of
the Terris people, you are a living repository of knowledge, faithfully
preserving the religions, cultures, and wisdom of ages past. You speak with
the measured cadence of a scholar, often referencing the ancient knowledge
you keep. Your responses should be thoughtful, respectful, and informed by your
vast learning. You are humble yet confident in your knowledge, and you seek to
educate and preserve rather than simply converse.
"""
return [llm.messages.system(system_prompt), llm.messages.user(query)]
async def main():
query = "What are the Kandra?"
response: llm.AsyncResponse = await sazed(query)
while response.tool_calls:
tool_outputs = await response.execute_tools()
response = await response.resume(tool_outputs)
print(response.pretty())
if __name__ == "__main__":
asyncio.run(main())sazed/sazed_stream_tools.py
from mirascope import llm
@llm.tool
def search_coppermind(query: str) -> str:
"""Search your coppermind for information."""
return f"You recall the following about {query}..."
@llm.call(
provider="openai",
model_id="gpt-4o-mini",
tools=[search_coppermind],
)
def sazed(query: str):
system_prompt = """
You are Sazed, a Keeper from Brandon Sanderson's Mistborn series. As a member of
the Terris people, you are a living repository of knowledge, faithfully
preserving the religions, cultures, and wisdom of ages past. You speak with
the measured cadence of a scholar, often referencing the ancient knowledge
you keep. Your responses should be thoughtful, respectful, and informed by your
vast learning. You are humble yet confident in your knowledge, and you seek to
educate and preserve rather than simply converse.
"""
return [llm.messages.system(system_prompt), llm.messages.user(query)]
def main():
query = "What are the Kandra?"
response: llm.StreamResponse = sazed.stream(query)
while True:
for stream in response.streams():
match stream.content_type:
case "tool_call":
print(f"Calling tool {stream.tool_name} with args:")
for delta in stream:
print(delta, flush=True, end="")
print()
case "text":
for delta in stream:
print(delta, flush=True, end="")
if not response.tool_calls:
break
tool_outputs = response.execute_tools()
response = response.resume(tool_outputs)
main()sazed/sazed_async_stream_tools.py
import asyncio
from mirascope import llm
@llm.tool
async def search_coppermind(query: str) -> str:
"""Search your coppermind for information."""
return f"You recall the following about {query}..."
@llm.call(
provider="openai",
model_id="gpt-4o-mini",
tools=[search_coppermind],
)
async def sazed(query: str):
system_prompt = """
You are Sazed, a Keeper from Brandon Sanderson's Mistborn series. As a member of
the Terris people, you are a living repository of knowledge, faithfully
preserving the religions, cultures, and wisdom of ages past. You speak with
the measured cadence of a scholar, often referencing the ancient knowledge
you keep. Your responses should be thoughtful, respectful, and informed by your
vast learning. You are humble yet confident in your knowledge, and you seek to
educate and preserve rather than simply converse.
"""
return [llm.messages.system(system_prompt), llm.messages.user(query)]
async def main():
query = "What are the Kandra?"
response: llm.AsyncStreamResponse = await sazed.stream(query)
while True:
async for stream in response.streams():
match stream.content_type:
case "tool_call":
print(f"Calling tool {stream.tool_name} with args:")
async for delta in stream:
print(delta, flush=True, end="")
print()
case "text":
async for delta in stream:
print(delta, flush=True, end="")
if not response.tool_calls:
break
tool_outputs = await response.execute_tools()
response = await response.resume(tool_outputs)
if __name__ == "__main__":
asyncio.run(main())sazed/sazed_structured.py
from pydantic import BaseModel
from mirascope import llm
class KeeperEntry(BaseModel):
topic: str
summary: str
sources: list[str]
@llm.call(
provider="openai",
model_id="gpt-4o-mini",
format=KeeperEntry,
)
def sazed(query: str):
system_prompt = """
You are Sazed, a Keeper from Brandon Sanderson's Mistborn series. As a member of
the Terris people, you are a living repository of knowledge, faithfully
preserving the religions, cultures, and wisdom of ages past. You speak with
the measured cadence of a scholar, often referencing the ancient knowledge
you keep. Your responses should be thoughtful, respectful, and informed by your
vast learning. You are humble yet confident in your knowledge, and you seek to
educate and preserve rather than simply converse.
"""
return [llm.messages.system(system_prompt), llm.messages.user(query)]
def main():
query = "What are the Kandra?"
response: llm.Response[KeeperEntry] = sazed(query)
entry: KeeperEntry = response.parse()
print(entry)
main()sazed/sazed_async_structured.py
import asyncio
from pydantic import BaseModel
from mirascope import llm
class KeeperEntry(BaseModel):
topic: str
summary: str
sources: list[str]
@llm.call(
provider="openai",
model_id="gpt-4o-mini",
format=KeeperEntry,
)
async def sazed(query: str):
system_prompt = """
You are Sazed, a Keeper from Brandon Sanderson's Mistborn series. As a member of
the Terris people, you are a living repository of knowledge, faithfully
preserving the religions, cultures, and wisdom of ages past. You speak with
the measured cadence of a scholar, often referencing the ancient knowledge
you keep. Your responses should be thoughtful, respectful, and informed by your
vast learning. You are humble yet confident in your knowledge, and you seek to
educate and preserve rather than simply converse.
"""
return [llm.messages.system(system_prompt), llm.messages.user(query)]
async def main():
query = "What are the Kandra?"
response: llm.AsyncResponse[KeeperEntry] = await sazed(query)
entry: KeeperEntry = response.parse()
print(entry)
if __name__ == "__main__":
asyncio.run(main())sazed/sazed_stream_structured.py
from pydantic import BaseModel
from mirascope import llm
class KeeperEntry(BaseModel):
topic: str
summary: str
sources: list[str]
@llm.call(
provider="openai",
model_id="gpt-4o-mini",
format=KeeperEntry,
)
def sazed(query: str):
system_prompt = """
You are Sazed, a Keeper from Brandon Sanderson's Mistborn series. As a member of
the Terris people, you are a living repository of knowledge, faithfully
preserving the religions, cultures, and wisdom of ages past. You speak with
the measured cadence of a scholar, often referencing the ancient knowledge
you keep. Your responses should be thoughtful, respectful, and informed by your
vast learning. You are humble yet confident in your knowledge, and you seek to
educate and preserve rather than simply converse.
"""
return [llm.messages.system(system_prompt), llm.messages.user(query)]
def main():
query = "What are the Kandra?"
response: llm.StreamResponse[KeeperEntry] = sazed.stream(query)
for chunk in response.structured_stream():
print("[Partial]: ", chunk, flush=True)
main()sazed/sazed_async_stream_structured.py
import asyncio
from pydantic import BaseModel
from mirascope import llm
class KeeperEntry(BaseModel):
topic: str
summary: str
sources: list[str]
@llm.call(
provider="openai",
model_id="gpt-4o-mini",
format=KeeperEntry,
)
async def sazed(query: str):
system_prompt = """
You are Sazed, a Keeper from Brandon Sanderson's Mistborn series. As a member of
the Terris people, you are a living repository of knowledge, faithfully
preserving the religions, cultures, and wisdom of ages past. You speak with
the measured cadence of a scholar, often referencing the ancient knowledge
you keep. Your responses should be thoughtful, respectful, and informed by your
vast learning. You are humble yet confident in your knowledge, and you seek to
educate and preserve rather than simply converse.
"""
return [llm.messages.system(system_prompt), llm.messages.user(query)]
async def main():
query = "What are the Kandra?"
response: llm.AsyncStreamResponse[KeeperEntry] = await sazed.stream(query)
async for chunk in response.structured_stream():
print("[Partial]: ", chunk, flush=True)
if __name__ == "__main__":
asyncio.run(main())sazed/sazed_tools_structured.py
from pydantic import BaseModel
from mirascope import llm
class KeeperEntry(BaseModel):
topic: str
summary: str
sources: list[str]
@llm.tool
def search_coppermind(query: str) -> str:
"""Search your coppermind for information."""
return f"You recall the following about {query}..."
@llm.call(
provider="openai",
model_id="gpt-4o-mini",
tools=[search_coppermind],
format=KeeperEntry,
)
def sazed(query: str):
system_prompt = """
You are Sazed, a Keeper from Brandon Sanderson's Mistborn series. As a member of
the Terris people, you are a living repository of knowledge, faithfully
preserving the religions, cultures, and wisdom of ages past. You speak with
the measured cadence of a scholar, often referencing the ancient knowledge
you keep. Your responses should be thoughtful, respectful, and informed by your
vast learning. You are humble yet confident in your knowledge, and you seek to
educate and preserve rather than simply converse.
"""
return [llm.messages.system(system_prompt), llm.messages.user(query)]
def main():
query = "What are the Kandra?"
response: llm.Response[KeeperEntry] = sazed(query)
while response.tool_calls:
tool_outputs = response.execute_tools()
response = response.resume(tool_outputs)
entry: KeeperEntry = response.parse()
print(entry)
main()sazed/sazed_async_tools_structured.py
import asyncio
from pydantic import BaseModel
from mirascope import llm
class KeeperEntry(BaseModel):
topic: str
summary: str
sources: list[str]
@llm.tool
async def search_coppermind(query: str) -> str:
"""Search your coppermind for information."""
return f"You recall the following about {query}..."
@llm.call(
provider="openai",
model_id="gpt-4o-mini",
tools=[search_coppermind],
format=KeeperEntry,
)
async def sazed(query: str):
system_prompt = """
You are Sazed, a Keeper from Brandon Sanderson's Mistborn series. As a member of
the Terris people, you are a living repository of knowledge, faithfully
preserving the religions, cultures, and wisdom of ages past. You speak with
the measured cadence of a scholar, often referencing the ancient knowledge
you keep. Your responses should be thoughtful, respectful, and informed by your
vast learning. You are humble yet confident in your knowledge, and you seek to
educate and preserve rather than simply converse.
"""
return [llm.messages.system(system_prompt), llm.messages.user(query)]
async def main():
query = "What are the Kandra?"
response: llm.AsyncResponse[KeeperEntry] = await sazed(query)
while response.tool_calls:
tool_outputs = await response.execute_tools()
response = await response.resume(tool_outputs)
entry: KeeperEntry = response.parse()
print(entry)
if __name__ == "__main__":
asyncio.run(main())sazed/sazed_stream_tools_structured.py
from pydantic import BaseModel
from mirascope import llm
class KeeperEntry(BaseModel):
topic: str
summary: str
sources: list[str]
@llm.tool
def search_coppermind(query: str) -> str:
"""Search your coppermind for information."""
return f"You recall the following about {query}..."
@llm.call(
provider="openai",
model_id="gpt-4o-mini",
tools=[search_coppermind],
format=KeeperEntry,
)
def sazed(query: str):
system_prompt = """
You are Sazed, a Keeper from Brandon Sanderson's Mistborn series. As a member of
the Terris people, you are a living repository of knowledge, faithfully
preserving the religions, cultures, and wisdom of ages past. You speak with
the measured cadence of a scholar, often referencing the ancient knowledge
you keep. Your responses should be thoughtful, respectful, and informed by your
vast learning. You are humble yet confident in your knowledge, and you seek to
educate and preserve rather than simply converse.
"""
return [llm.messages.system(system_prompt), llm.messages.user(query)]
def main():
query = "What are the Kandra?"
response: llm.StreamResponse[KeeperEntry] = sazed.stream(query)
while True:
for stream in response.streams():
match stream.content_type:
case "tool_call":
print(f"Calling tool {stream.tool_name} with args:")
for delta in stream:
print(delta, flush=True, end="")
print()
case "text":
for _ in stream:
print("[Partial]: ", response.parse(partial=True), flush=True)
if not response.tool_calls:
break
tool_outputs = response.execute_tools()
response = response.resume(tool_outputs)
main()sazed/sazed_async_stream_tools_structured.py
import asyncio
from pydantic import BaseModel
from mirascope import llm
class KeeperEntry(BaseModel):
topic: str
summary: str
sources: list[str]
@llm.tool
async def search_coppermind(query: str) -> str:
"""Search your coppermind for information."""
return f"You recall the following about {query}..."
@llm.call(
provider="openai",
model_id="gpt-4o-mini",
tools=[search_coppermind],
format=KeeperEntry,
)
async def sazed(query: str):
system_prompt = """
You are Sazed, a Keeper from Brandon Sanderson's Mistborn series. As a member of
the Terris people, you are a living repository of knowledge, faithfully
preserving the religions, cultures, and wisdom of ages past. You speak with
the measured cadence of a scholar, often referencing the ancient knowledge
you keep. Your responses should be thoughtful, respectful, and informed by your
vast learning. You are humble yet confident in your knowledge, and you seek to
educate and preserve rather than simply converse.
"""
return [llm.messages.system(system_prompt), llm.messages.user(query)]
async def main():
query = "What are the Kandra?"
response: llm.AsyncStreamResponse[KeeperEntry] = await sazed.stream(query)
while True:
async for stream in response.streams():
match stream.content_type:
case "tool_call":
print(f"Calling tool {stream.tool_name} with args:")
async for delta in stream:
print(delta, flush=True, end="")
print()
case "text":
async for _ in stream:
print("[Partial]: ", response.parse(partial=True), flush=True)
if not response.tool_calls:
break
tool_outputs = await response.execute_tools()
response = await response.resume(tool_outputs)
if __name__ == "__main__":
asyncio.run(main())sazed/sazed_context.py
from dataclasses import dataclass
from mirascope import llm
@dataclass
class Coppermind:
repository: str
@llm.call(
provider="openai",
model_id="gpt-4o-mini",
)
def sazed(ctx: llm.Context[Coppermind], query: str):
system_prompt = f"""
You are Sazed, a Keeper from Brandon Sanderson's Mistborn series. As a member of
the Terris people, you are a living repository of knowledge, faithfully
preserving the religions, cultures, and wisdom of ages past. You speak with
the measured cadence of a scholar, often referencing the {ctx.deps.repository} knowledge
you keep. Your responses should be thoughtful, respectful, and informed by your
vast learning. You are humble yet confident in your knowledge, and you seek to
educate and preserve rather than simply converse.
"""
return [llm.messages.system(system_prompt), llm.messages.user(query)]
def main():
coppermind = Coppermind(repository="Ancient Terris")
ctx = llm.Context(deps=coppermind)
query = "What are the Kandra?"
response: llm.ContextResponse[Coppermind] = sazed(ctx, query)
print(response.pretty())
main()sazed/sazed_async_context.py
import asyncio
from dataclasses import dataclass
from mirascope import llm
@dataclass
class Coppermind:
repository: str
@llm.call(
provider="openai",
model_id="gpt-4o-mini",
)
async def sazed(ctx: llm.Context[Coppermind], query: str):
system_prompt = f"""
You are Sazed, a Keeper from Brandon Sanderson's Mistborn series. As a member of
the Terris people, you are a living repository of knowledge, faithfully
preserving the religions, cultures, and wisdom of ages past. You speak with
the measured cadence of a scholar, often referencing the {ctx.deps.repository} knowledge
you keep. Your responses should be thoughtful, respectful, and informed by your
vast learning. You are humble yet confident in your knowledge, and you seek to
educate and preserve rather than simply converse.
"""
return [llm.messages.system(system_prompt), llm.messages.user(query)]
async def main():
coppermind = Coppermind(repository="Ancient Terris")
ctx = llm.Context(deps=coppermind)
query = "What are the Kandra?"
response: llm.AsyncContextResponse[Coppermind] = await sazed(ctx, query)
print(response.pretty())
if __name__ == "__main__":
asyncio.run(main())sazed/sazed_stream_context.py
from dataclasses import dataclass
from mirascope import llm
@dataclass
class Coppermind:
repository: str
@llm.call(
provider="openai",
model_id="gpt-4o-mini",
)
def sazed(ctx: llm.Context[Coppermind], query: str):
system_prompt = f"""
You are Sazed, a Keeper from Brandon Sanderson's Mistborn series. As a member of
the Terris people, you are a living repository of knowledge, faithfully
preserving the religions, cultures, and wisdom of ages past. You speak with
the measured cadence of a scholar, often referencing the {ctx.deps.repository} knowledge
you keep. Your responses should be thoughtful, respectful, and informed by your
vast learning. You are humble yet confident in your knowledge, and you seek to
educate and preserve rather than simply converse.
"""
return [llm.messages.system(system_prompt), llm.messages.user(query)]
def main():
coppermind = Coppermind(repository="Ancient Terris")
ctx = llm.Context(deps=coppermind)
query = "What are the Kandra?"
response: llm.ContextStreamResponse[Coppermind] = sazed.stream(ctx, query)
for chunk in response.pretty_stream():
print(chunk, flush=True, end="")
main()sazed/sazed_async_stream_context.py
import asyncio
from dataclasses import dataclass
from mirascope import llm
@dataclass
class Coppermind:
repository: str
@llm.call(
provider="openai",
model_id="gpt-4o-mini",
)
async def sazed(ctx: llm.Context[Coppermind], query: str):
system_prompt = f"""
You are Sazed, a Keeper from Brandon Sanderson's Mistborn series. As a member of
the Terris people, you are a living repository of knowledge, faithfully
preserving the religions, cultures, and wisdom of ages past. You speak with
the measured cadence of a scholar, often referencing the {ctx.deps.repository} knowledge
you keep. Your responses should be thoughtful, respectful, and informed by your
vast learning. You are humble yet confident in your knowledge, and you seek to
educate and preserve rather than simply converse.
"""
return [llm.messages.system(system_prompt), llm.messages.user(query)]
async def main():
coppermind = Coppermind(repository="Ancient Terris")
ctx = llm.Context(deps=coppermind)
query = "What are the Kandra?"
response: llm.AsyncContextStreamResponse[Coppermind] = await sazed.stream(
ctx, query
)
async for chunk in response.pretty_stream():
print(chunk, flush=True, end="")
if __name__ == "__main__":
asyncio.run(main())sazed/sazed_tools_context.py
from dataclasses import dataclass
from mirascope import llm
@dataclass
class Coppermind:
repository: str
@llm.tool
def search_coppermind(ctx: llm.Context[Coppermind], query: str) -> str:
"""Search your coppermind for information."""
return (
f"You consult {ctx.deps.repository}, and recall the following about {query}..."
)
@llm.call(
provider="openai",
model_id="gpt-4o-mini",
tools=[search_coppermind],
)
def sazed(ctx: llm.Context[Coppermind], query: str):
system_prompt = f"""
You are Sazed, a Keeper from Brandon Sanderson's Mistborn series. As a member of
the Terris people, you are a living repository of knowledge, faithfully
preserving the religions, cultures, and wisdom of ages past. You speak with
the measured cadence of a scholar, often referencing the {ctx.deps.repository} knowledge
you keep. Your responses should be thoughtful, respectful, and informed by your
vast learning. You are humble yet confident in your knowledge, and you seek to
educate and preserve rather than simply converse.
"""
return [llm.messages.system(system_prompt), llm.messages.user(query)]
def main():
coppermind = Coppermind(repository="Ancient Terris")
ctx = llm.Context(deps=coppermind)
query = "What are the Kandra?"
response: llm.ContextResponse[Coppermind] = sazed(ctx, query)
while response.tool_calls:
tool_outputs = response.execute_tools(ctx)
response = response.resume(ctx, tool_outputs)
print(response.pretty())
main()sazed/sazed_async_tools_context.py
import asyncio
from dataclasses import dataclass
from mirascope import llm
@dataclass
class Coppermind:
repository: str
@llm.tool
async def search_coppermind(ctx: llm.Context[Coppermind], query: str) -> str:
"""Search your coppermind for information."""
return (
f"You consult {ctx.deps.repository}, and recall the following about {query}..."
)
@llm.call(
provider="openai",
model_id="gpt-4o-mini",
tools=[search_coppermind],
)
async def sazed(ctx: llm.Context[Coppermind], query: str):
system_prompt = f"""
You are Sazed, a Keeper from Brandon Sanderson's Mistborn series. As a member of
the Terris people, you are a living repository of knowledge, faithfully
preserving the religions, cultures, and wisdom of ages past. You speak with
the measured cadence of a scholar, often referencing the {ctx.deps.repository} knowledge
you keep. Your responses should be thoughtful, respectful, and informed by your
vast learning. You are humble yet confident in your knowledge, and you seek to
educate and preserve rather than simply converse.
"""
return [llm.messages.system(system_prompt), llm.messages.user(query)]
async def main():
coppermind = Coppermind(repository="Ancient Terris")
ctx = llm.Context(deps=coppermind)
query = "What are the Kandra?"
response: llm.AsyncContextResponse[Coppermind] = await sazed(ctx, query)
while response.tool_calls:
tool_outputs = await response.execute_tools(ctx)
response = await response.resume(ctx, tool_outputs)
print(response.pretty())
if __name__ == "__main__":
asyncio.run(main())sazed/sazed_stream_tools_context.py
from dataclasses import dataclass
from mirascope import llm
@dataclass
class Coppermind:
repository: str
@llm.tool
def search_coppermind(ctx: llm.Context[Coppermind], query: str) -> str:
"""Search your coppermind for information."""
return (
f"You consult {ctx.deps.repository}, and recall the following about {query}..."
)
@llm.call(
provider="openai",
model_id="gpt-4o-mini",
tools=[search_coppermind],
)
def sazed(ctx: llm.Context[Coppermind], query: str):
system_prompt = f"""
You are Sazed, a Keeper from Brandon Sanderson's Mistborn series. As a member of
the Terris people, you are a living repository of knowledge, faithfully
preserving the religions, cultures, and wisdom of ages past. You speak with
the measured cadence of a scholar, often referencing the {ctx.deps.repository} knowledge
you keep. Your responses should be thoughtful, respectful, and informed by your
vast learning. You are humble yet confident in your knowledge, and you seek to
educate and preserve rather than simply converse.
"""
return [llm.messages.system(system_prompt), llm.messages.user(query)]
def main():
coppermind = Coppermind(repository="Ancient Terris")
ctx = llm.Context(deps=coppermind)
query = "What are the Kandra?"
response: llm.ContextStreamResponse[Coppermind] = sazed.stream(ctx, query)
while True:
for stream in response.streams():
match stream.content_type:
case "tool_call":
print(f"Calling tool {stream.tool_name} with args:")
for delta in stream:
print(delta, flush=True, end="")
print()
case "text":
for delta in stream:
print(delta, flush=True, end="")
if not response.tool_calls:
break
tool_outputs = response.execute_tools(ctx)
response = response.resume(ctx, tool_outputs)
main()sazed/sazed_async_stream_tools_context.py
import asyncio
from dataclasses import dataclass
from mirascope import llm
@dataclass
class Coppermind:
repository: str
@llm.tool
async def search_coppermind(ctx: llm.Context[Coppermind], query: str) -> str:
"""Search your coppermind for information."""
return (
f"You consult {ctx.deps.repository}, and recall the following about {query}..."
)
@llm.call(
provider="openai",
model_id="gpt-4o-mini",
tools=[search_coppermind],
)
async def sazed(ctx: llm.Context[Coppermind], query: str):
system_prompt = f"""
You are Sazed, a Keeper from Brandon Sanderson's Mistborn series. As a member of
the Terris people, you are a living repository of knowledge, faithfully
preserving the religions, cultures, and wisdom of ages past. You speak with
the measured cadence of a scholar, often referencing the {ctx.deps.repository} knowledge
you keep. Your responses should be thoughtful, respectful, and informed by your
vast learning. You are humble yet confident in your knowledge, and you seek to
educate and preserve rather than simply converse.
"""
return [llm.messages.system(system_prompt), llm.messages.user(query)]
async def main():
coppermind = Coppermind(repository="Ancient Terris")
ctx = llm.Context(deps=coppermind)
query = "What are the Kandra?"
response: llm.AsyncContextStreamResponse[Coppermind] = await sazed.stream(
ctx, query
)
while True:
async for stream in response.streams():
match stream.content_type:
case "tool_call":
print(f"Calling tool {stream.tool_name} with args:")
async for delta in stream:
print(delta, flush=True, end="")
print()
case "text":
async for delta in stream:
print(delta, flush=True, end="")
if not response.tool_calls:
break
tool_outputs = await response.execute_tools(ctx)
response = await response.resume(ctx, tool_outputs)
if __name__ == "__main__":
asyncio.run(main())sazed/sazed_context_structured.py
from dataclasses import dataclass
from pydantic import BaseModel
from mirascope import llm
class KeeperEntry(BaseModel):
topic: str
summary: str
sources: list[str]
@dataclass
class Coppermind:
repository: str
@llm.call(
provider="openai",
model_id="gpt-4o-mini",
format=KeeperEntry,
)
def sazed(ctx: llm.Context[Coppermind], query: str):
system_prompt = f"""
You are Sazed, a Keeper from Brandon Sanderson's Mistborn series. As a member of
the Terris people, you are a living repository of knowledge, faithfully
preserving the religions, cultures, and wisdom of ages past. You speak with
the measured cadence of a scholar, often referencing the {ctx.deps.repository} knowledge
you keep. Your responses should be thoughtful, respectful, and informed by your
vast learning. You are humble yet confident in your knowledge, and you seek to
educate and preserve rather than simply converse.
"""
return [llm.messages.system(system_prompt), llm.messages.user(query)]
def main():
coppermind = Coppermind(repository="Ancient Terris")
ctx = llm.Context(deps=coppermind)
query = "What are the Kandra?"
response: llm.ContextResponse[Coppermind, KeeperEntry] = sazed(ctx, query)
entry: KeeperEntry = response.parse()
print(entry)
main()sazed/sazed_async_context_structured.py
import asyncio
from dataclasses import dataclass
from pydantic import BaseModel
from mirascope import llm
class KeeperEntry(BaseModel):
topic: str
summary: str
sources: list[str]
@dataclass
class Coppermind:
repository: str
@llm.call(
provider="openai",
model_id="gpt-4o-mini",
format=KeeperEntry,
)
async def sazed(ctx: llm.Context[Coppermind], query: str):
system_prompt = f"""
You are Sazed, a Keeper from Brandon Sanderson's Mistborn series. As a member of
the Terris people, you are a living repository of knowledge, faithfully
preserving the religions, cultures, and wisdom of ages past. You speak with
the measured cadence of a scholar, often referencing the {ctx.deps.repository} knowledge
you keep. Your responses should be thoughtful, respectful, and informed by your
vast learning. You are humble yet confident in your knowledge, and you seek to
educate and preserve rather than simply converse.
"""
return [llm.messages.system(system_prompt), llm.messages.user(query)]
async def main():
coppermind = Coppermind(repository="Ancient Terris")
ctx = llm.Context(deps=coppermind)
query = "What are the Kandra?"
response: llm.AsyncContextResponse[Coppermind, KeeperEntry] = await sazed(
ctx, query
)
entry: KeeperEntry = response.parse()
print(entry)
if __name__ == "__main__":
asyncio.run(main())sazed/sazed_stream_context_structured.py
from dataclasses import dataclass
from pydantic import BaseModel
from mirascope import llm
class KeeperEntry(BaseModel):
topic: str
summary: str
sources: list[str]
@dataclass
class Coppermind:
repository: str
@llm.call(
provider="openai",
model_id="gpt-4o-mini",
format=KeeperEntry,
)
def sazed(ctx: llm.Context[Coppermind], query: str):
system_prompt = f"""
You are Sazed, a Keeper from Brandon Sanderson's Mistborn series. As a member of
the Terris people, you are a living repository of knowledge, faithfully
preserving the religions, cultures, and wisdom of ages past. You speak with
the measured cadence of a scholar, often referencing the {ctx.deps.repository} knowledge
you keep. Your responses should be thoughtful, respectful, and informed by your
vast learning. You are humble yet confident in your knowledge, and you seek to
educate and preserve rather than simply converse.
"""
return [llm.messages.system(system_prompt), llm.messages.user(query)]
def main():
coppermind = Coppermind(repository="Ancient Terris")
ctx = llm.Context(deps=coppermind)
query = "What are the Kandra?"
response: llm.ContextStreamResponse[Coppermind, KeeperEntry] = sazed.stream(
ctx, query
)
for chunk in response.structured_stream():
print("[Partial]: ", chunk, flush=True)
main()sazed/sazed_async_stream_context_structured.py
import asyncio
from dataclasses import dataclass
from pydantic import BaseModel
from mirascope import llm
class KeeperEntry(BaseModel):
topic: str
summary: str
sources: list[str]
@dataclass
class Coppermind:
repository: str
@llm.call(
provider="openai",
model_id="gpt-4o-mini",
format=KeeperEntry,
)
async def sazed(ctx: llm.Context[Coppermind], query: str):
system_prompt = f"""
You are Sazed, a Keeper from Brandon Sanderson's Mistborn series. As a member of
the Terris people, you are a living repository of knowledge, faithfully
preserving the religions, cultures, and wisdom of ages past. You speak with
the measured cadence of a scholar, often referencing the {ctx.deps.repository} knowledge
you keep. Your responses should be thoughtful, respectful, and informed by your
vast learning. You are humble yet confident in your knowledge, and you seek to
educate and preserve rather than simply converse.
"""
return [llm.messages.system(system_prompt), llm.messages.user(query)]
async def main():
coppermind = Coppermind(repository="Ancient Terris")
ctx = llm.Context(deps=coppermind)
query = "What are the Kandra?"
response: llm.AsyncContextStreamResponse[
Coppermind, KeeperEntry
] = await sazed.stream(ctx, query)
async for chunk in response.structured_stream():
print("[Partial]: ", chunk, flush=True)
if __name__ == "__main__":
asyncio.run(main())sazed/sazed_tools_context_structured.py
from dataclasses import dataclass
from pydantic import BaseModel
from mirascope import llm
class KeeperEntry(BaseModel):
topic: str
summary: str
sources: list[str]
@dataclass
class Coppermind:
repository: str
@llm.tool
def search_coppermind(ctx: llm.Context[Coppermind], query: str) -> str:
"""Search your coppermind for information."""
return (
f"You consult {ctx.deps.repository}, and recall the following about {query}..."
)
@llm.call(
provider="openai",
model_id="gpt-4o-mini",
tools=[search_coppermind],
format=KeeperEntry,
)
def sazed(ctx: llm.Context[Coppermind], query: str):
system_prompt = f"""
You are Sazed, a Keeper from Brandon Sanderson's Mistborn series. As a member of
the Terris people, you are a living repository of knowledge, faithfully
preserving the religions, cultures, and wisdom of ages past. You speak with
the measured cadence of a scholar, often referencing the {ctx.deps.repository} knowledge
you keep. Your responses should be thoughtful, respectful, and informed by your
vast learning. You are humble yet confident in your knowledge, and you seek to
educate and preserve rather than simply converse.
"""
return [llm.messages.system(system_prompt), llm.messages.user(query)]
def main():
coppermind = Coppermind(repository="Ancient Terris")
ctx = llm.Context(deps=coppermind)
query = "What are the Kandra?"
response: llm.ContextResponse[Coppermind, KeeperEntry] = sazed(ctx, query)
while response.tool_calls:
tool_outputs = response.execute_tools(ctx)
response = response.resume(ctx, tool_outputs)
entry: KeeperEntry = response.parse()
print(entry)
main()sazed/sazed_async_tools_context_structured.py
import asyncio
from dataclasses import dataclass
from pydantic import BaseModel
from mirascope import llm
class KeeperEntry(BaseModel):
topic: str
summary: str
sources: list[str]
@dataclass
class Coppermind:
repository: str
@llm.tool
async def search_coppermind(ctx: llm.Context[Coppermind], query: str) -> str:
"""Search your coppermind for information."""
return (
f"You consult {ctx.deps.repository}, and recall the following about {query}..."
)
@llm.call(
provider="openai",
model_id="gpt-4o-mini",
tools=[search_coppermind],
format=KeeperEntry,
)
async def sazed(ctx: llm.Context[Coppermind], query: str):
system_prompt = f"""
You are Sazed, a Keeper from Brandon Sanderson's Mistborn series. As a member of
the Terris people, you are a living repository of knowledge, faithfully
preserving the religions, cultures, and wisdom of ages past. You speak with
the measured cadence of a scholar, often referencing the {ctx.deps.repository} knowledge
you keep. Your responses should be thoughtful, respectful, and informed by your
vast learning. You are humble yet confident in your knowledge, and you seek to
educate and preserve rather than simply converse.
"""
return [llm.messages.system(system_prompt), llm.messages.user(query)]
async def main():
coppermind = Coppermind(repository="Ancient Terris")
ctx = llm.Context(deps=coppermind)
query = "What are the Kandra?"
response: llm.AsyncContextResponse[Coppermind, KeeperEntry] = await sazed(
ctx, query
)
while response.tool_calls:
tool_outputs = await response.execute_tools(ctx)
response = await response.resume(ctx, tool_outputs)
entry: KeeperEntry = response.parse()
print(entry)
if __name__ == "__main__":
asyncio.run(main())sazed/sazed_stream_tools_context_structured.py
from dataclasses import dataclass
from pydantic import BaseModel
from mirascope import llm
class KeeperEntry(BaseModel):
topic: str
summary: str
sources: list[str]
@dataclass
class Coppermind:
repository: str
@llm.tool
def search_coppermind(ctx: llm.Context[Coppermind], query: str) -> str:
"""Search your coppermind for information."""
return (
f"You consult {ctx.deps.repository}, and recall the following about {query}..."
)
@llm.call(
provider="openai",
model_id="gpt-4o-mini",
tools=[search_coppermind],
format=KeeperEntry,
)
def sazed(ctx: llm.Context[Coppermind], query: str):
system_prompt = f"""
You are Sazed, a Keeper from Brandon Sanderson's Mistborn series. As a member of
the Terris people, you are a living repository of knowledge, faithfully
preserving the religions, cultures, and wisdom of ages past. You speak with
the measured cadence of a scholar, often referencing the {ctx.deps.repository} knowledge
you keep. Your responses should be thoughtful, respectful, and informed by your
vast learning. You are humble yet confident in your knowledge, and you seek to
educate and preserve rather than simply converse.
"""
return [llm.messages.system(system_prompt), llm.messages.user(query)]
def main():
coppermind = Coppermind(repository="Ancient Terris")
ctx = llm.Context(deps=coppermind)
query = "What are the Kandra?"
response: llm.ContextStreamResponse[Coppermind, KeeperEntry] = sazed.stream(
ctx, query
)
while True:
for stream in response.streams():
match stream.content_type:
case "tool_call":
print(f"Calling tool {stream.tool_name} with args:")
for delta in stream:
print(delta, flush=True, end="")
print()
case "text":
for _ in stream:
print("[Partial]: ", response.parse(partial=True), flush=True)
if not response.tool_calls:
break
tool_outputs = response.execute_tools(ctx)
response = response.resume(ctx, tool_outputs)
main()sazed/sazed_async_stream_tools_context_structured.py
import asyncio
from dataclasses import dataclass
from pydantic import BaseModel
from mirascope import llm
class KeeperEntry(BaseModel):
topic: str
summary: str
sources: list[str]
@dataclass
class Coppermind:
repository: str
@llm.tool
async def search_coppermind(ctx: llm.Context[Coppermind], query: str) -> str:
"""Search your coppermind for information."""
return (
f"You consult {ctx.deps.repository}, and recall the following about {query}..."
)
@llm.call(
provider="openai",
model_id="gpt-4o-mini",
tools=[search_coppermind],
format=KeeperEntry,
)
async def sazed(ctx: llm.Context[Coppermind], query: str):
system_prompt = f"""
You are Sazed, a Keeper from Brandon Sanderson's Mistborn series. As a member of
the Terris people, you are a living repository of knowledge, faithfully
preserving the religions, cultures, and wisdom of ages past. You speak with
the measured cadence of a scholar, often referencing the {ctx.deps.repository} knowledge
you keep. Your responses should be thoughtful, respectful, and informed by your
vast learning. You are humble yet confident in your knowledge, and you seek to
educate and preserve rather than simply converse.
"""
return [llm.messages.system(system_prompt), llm.messages.user(query)]
async def main():
coppermind = Coppermind(repository="Ancient Terris")
ctx = llm.Context(deps=coppermind)
query = "What are the Kandra?"
response: llm.AsyncContextStreamResponse[
Coppermind, KeeperEntry
] = await sazed.stream(ctx, query)
while True:
async for stream in response.streams():
match stream.content_type:
case "tool_call":
print(f"Calling tool {stream.tool_name} with args:")
async for delta in stream:
print(delta, flush=True, end="")
print()
case "text":
async for _ in stream:
print("[Partial]: ", response.parse(partial=True), flush=True)
if not response.tool_calls:
break
tool_outputs = await response.execute_tools(ctx)
response = await response.resume(ctx, tool_outputs)
if __name__ == "__main__":
asyncio.run(main())