mirror of
https://github.com/Significant-Gravitas/AutoGPT.git
synced 2026-04-08 03:00:28 -04:00
feat(backend/blocks): Add GPT-5.1 and GPT-5.1-codex (#11406)
This pr adds the latest gpt-5.1 and gpt-5.1-codex llm's from openai, as well as update the price of the gpt-5-chat model https://platform.openai.com/docs/models/gpt-5.1 https://platform.openai.com/docs/models/gpt-5.1-codex I have also had to add a new codex block as it uses a different openai API and has other options the main llm's dont use <img width="231" height="755" alt="image" src="https://github.com/user-attachments/assets/a4056633-7b0f-446f-ae86-d7755c5b88ec" /> #### For code changes: - [x] I have clearly listed my changes in the PR description - [x] I have made a test plan - [x] I have tested my changes according to the test plan: <!-- Put your test plan here: --> - [x] Test the latest gpt-5.1 llm - [x] Test the latest gpt-5.1-codex block --------- Co-authored-by: Zamil Majdy <zamil.majdy@agpt.co> Co-authored-by: Claude <noreply@anthropic.com>
This commit is contained in:
Bently
224
autogpt_platform/backend/backend/blocks/codex.py
Normal file
224
autogpt_platform/backend/backend/blocks/codex.py
Normal file
@@ -0,0 +1,224 @@
|
||||
from dataclasses import dataclass
|
||||
from enum import Enum
|
||||
from typing import Any, Literal
|
||||
|
||||
from openai import AsyncOpenAI
|
||||
from openai.types.responses import Response as OpenAIResponse
|
||||
from pydantic import SecretStr
|
||||
|
||||
from backend.data.block import (
|
||||
Block,
|
||||
BlockCategory,
|
||||
BlockOutput,
|
||||
BlockSchemaInput,
|
||||
BlockSchemaOutput,
|
||||
)
|
||||
from backend.data.model import (
|
||||
APIKeyCredentials,
|
||||
CredentialsField,
|
||||
CredentialsMetaInput,
|
||||
NodeExecutionStats,
|
||||
SchemaField,
|
||||
)
|
||||
from backend.integrations.providers import ProviderName
|
||||
|
||||
|
||||
@dataclass
|
||||
class CodexCallResult:
|
||||
"""Structured response returned by Codex invocations."""
|
||||
|
||||
response: str
|
||||
reasoning: str
|
||||
response_id: str
|
||||
|
||||
|
||||
class CodexModel(str, Enum):
|
||||
"""Codex-capable OpenAI models."""
|
||||
|
||||
GPT5_1_CODEX = "gpt-5.1-codex"
|
||||
|
||||
|
||||
class CodexReasoningEffort(str, Enum):
|
||||
"""Configuration for the Responses API reasoning effort."""
|
||||
|
||||
NONE = "none"
|
||||
LOW = "low"
|
||||
MEDIUM = "medium"
|
||||
HIGH = "high"
|
||||
|
||||
|
||||
CodexCredentials = CredentialsMetaInput[
|
||||
Literal[ProviderName.OPENAI], Literal["api_key"]
|
||||
]
|
||||
|
||||
TEST_CREDENTIALS = APIKeyCredentials(
|
||||
id="e2fcb203-3f2d-4ad4-a344-8df3bc7db36b",
|
||||
provider="openai",
|
||||
api_key=SecretStr("mock-openai-api-key"),
|
||||
title="Mock OpenAI API key",
|
||||
expires_at=None,
|
||||
)
|
||||
TEST_CREDENTIALS_INPUT = {
|
||||
"provider": TEST_CREDENTIALS.provider,
|
||||
"id": TEST_CREDENTIALS.id,
|
||||
"type": TEST_CREDENTIALS.type,
|
||||
"title": TEST_CREDENTIALS.title,
|
||||
}
|
||||
|
||||
|
||||
def CodexCredentialsField() -> CodexCredentials:
|
||||
return CredentialsField(
|
||||
description="OpenAI API key with access to Codex models (Responses API).",
|
||||
)
|
||||
|
||||
|
||||
class CodeGenerationBlock(Block):
|
||||
"""Block that talks to Codex models via the OpenAI Responses API."""
|
||||
|
||||
class Input(BlockSchemaInput):
|
||||
prompt: str = SchemaField(
|
||||
description="Primary coding request passed to the Codex model.",
|
||||
placeholder="Generate a Python function that reverses a list.",
|
||||
)
|
||||
system_prompt: str = SchemaField(
|
||||
title="System Prompt",
|
||||
default=(
|
||||
"You are Codex, an elite software engineer. "
|
||||
"Favor concise, working code and highlight important caveats."
|
||||
),
|
||||
description="Optional instructions injected via the Responses API instructions field.",
|
||||
advanced=True,
|
||||
)
|
||||
model: CodexModel = SchemaField(
|
||||
title="Codex Model",
|
||||
default=CodexModel.GPT5_1_CODEX,
|
||||
description="Codex-optimized model served via the Responses API.",
|
||||
advanced=False,
|
||||
)
|
||||
reasoning_effort: CodexReasoningEffort = SchemaField(
|
||||
title="Reasoning Effort",
|
||||
default=CodexReasoningEffort.MEDIUM,
|
||||
description="Controls the Responses API reasoning budget. Select 'none' to skip reasoning configs.",
|
||||
advanced=True,
|
||||
)
|
||||
max_output_tokens: int | None = SchemaField(
|
||||
title="Max Output Tokens",
|
||||
default=2048,
|
||||
description="Upper bound for generated tokens (hard limit 128,000). Leave blank to let OpenAI decide.",
|
||||
advanced=True,
|
||||
)
|
||||
credentials: CodexCredentials = CodexCredentialsField()
|
||||
|
||||
class Output(BlockSchemaOutput):
|
||||
response: str = SchemaField(
|
||||
description="Code-focused response returned by the Codex model."
|
||||
)
|
||||
reasoning: str = SchemaField(
|
||||
description="Reasoning summary returned by the model, if available.",
|
||||
default="",
|
||||
)
|
||||
response_id: str = SchemaField(
|
||||
description="ID of the Responses API call for auditing/debugging.",
|
||||
default="",
|
||||
)
|
||||
|
||||
def __init__(self):
|
||||
super().__init__(
|
||||
id="86a2a099-30df-47b4-b7e4-34ae5f83e0d5",
|
||||
description="Generate or refactor code using OpenAI's Codex (Responses API).",
|
||||
categories={BlockCategory.AI, BlockCategory.DEVELOPER_TOOLS},
|
||||
input_schema=CodeGenerationBlock.Input,
|
||||
output_schema=CodeGenerationBlock.Output,
|
||||
test_input=[
|
||||
{
|
||||
"prompt": "Write a TypeScript function that deduplicates an array.",
|
||||
"credentials": TEST_CREDENTIALS_INPUT,
|
||||
}
|
||||
],
|
||||
test_output=[
|
||||
("response", str),
|
||||
("reasoning", str),
|
||||
("response_id", str),
|
||||
],
|
||||
test_mock={
|
||||
"call_codex": lambda *_args, **_kwargs: CodexCallResult(
|
||||
response="function dedupe<T>(items: T[]): T[] { return [...new Set(items)]; }",
|
||||
reasoning="Used Set to remove duplicates in O(n).",
|
||||
response_id="resp_test",
|
||||
)
|
||||
},
|
||||
test_credentials=TEST_CREDENTIALS,
|
||||
)
|
||||
self.execution_stats = NodeExecutionStats()
|
||||
|
||||
async def call_codex(
|
||||
self,
|
||||
*,
|
||||
credentials: APIKeyCredentials,
|
||||
model: CodexModel,
|
||||
prompt: str,
|
||||
system_prompt: str,
|
||||
max_output_tokens: int | None,
|
||||
reasoning_effort: CodexReasoningEffort,
|
||||
) -> CodexCallResult:
|
||||
"""Invoke the OpenAI Responses API."""
|
||||
client = AsyncOpenAI(api_key=credentials.api_key.get_secret_value())
|
||||
|
||||
request_payload: dict[str, Any] = {
|
||||
"model": model.value,
|
||||
"input": prompt,
|
||||
}
|
||||
if system_prompt:
|
||||
request_payload["instructions"] = system_prompt
|
||||
if max_output_tokens is not None:
|
||||
request_payload["max_output_tokens"] = max_output_tokens
|
||||
if reasoning_effort != CodexReasoningEffort.NONE:
|
||||
request_payload["reasoning"] = {"effort": reasoning_effort.value}
|
||||
|
||||
response = await client.responses.create(**request_payload)
|
||||
if not isinstance(response, OpenAIResponse):
|
||||
raise TypeError(f"Expected OpenAIResponse, got {type(response).__name__}")
|
||||
|
||||
# Extract data directly from typed response
|
||||
text_output = response.output_text or ""
|
||||
reasoning_summary = (
|
||||
str(response.reasoning.summary)
|
||||
if response.reasoning and response.reasoning.summary
|
||||
else ""
|
||||
)
|
||||
response_id = response.id or ""
|
||||
|
||||
# Update usage stats
|
||||
self.execution_stats.input_token_count = (
|
||||
response.usage.input_tokens if response.usage else 0
|
||||
)
|
||||
self.execution_stats.output_token_count = (
|
||||
response.usage.output_tokens if response.usage else 0
|
||||
)
|
||||
self.execution_stats.llm_call_count += 1
|
||||
|
||||
return CodexCallResult(
|
||||
response=text_output,
|
||||
reasoning=reasoning_summary,
|
||||
response_id=response_id,
|
||||
)
|
||||
|
||||
async def run(
|
||||
self,
|
||||
input_data: Input,
|
||||
*,
|
||||
credentials: APIKeyCredentials,
|
||||
**_kwargs,
|
||||
) -> BlockOutput:
|
||||
result = await self.call_codex(
|
||||
credentials=credentials,
|
||||
model=input_data.model,
|
||||
prompt=input_data.prompt,
|
||||
system_prompt=input_data.system_prompt,
|
||||
max_output_tokens=input_data.max_output_tokens,
|
||||
reasoning_effort=input_data.reasoning_effort,
|
||||
)
|
||||
|
||||
yield "response", result.response
|
||||
yield "reasoning", result.reasoning
|
||||
yield "response_id", result.response_id
|
||||
@@ -93,6 +93,7 @@ class LlmModel(str, Enum, metaclass=LlmModelMeta):
|
||||
O1_MINI = "o1-mini"
|
||||
# GPT-5 models
|
||||
GPT5 = "gpt-5-2025-08-07"
|
||||
GPT5_1 = "gpt-5.1-2025-11-13"
|
||||
GPT5_MINI = "gpt-5-mini-2025-08-07"
|
||||
GPT5_NANO = "gpt-5-nano-2025-08-07"
|
||||
GPT5_CHAT = "gpt-5-chat-latest"
|
||||
@@ -194,6 +195,7 @@ MODEL_METADATA = {
|
||||
LlmModel.O1_MINI: ModelMetadata("openai", 128000, 65536), # o1-mini-2024-09-12
|
||||
# GPT-5 models
|
||||
LlmModel.GPT5: ModelMetadata("openai", 400000, 128000),
|
||||
LlmModel.GPT5_1: ModelMetadata("openai", 400000, 128000),
|
||||
LlmModel.GPT5_MINI: ModelMetadata("openai", 400000, 128000),
|
||||
LlmModel.GPT5_NANO: ModelMetadata("openai", 400000, 128000),
|
||||
LlmModel.GPT5_CHAT: ModelMetadata("openai", 400000, 16384),
|
||||
|
||||
@@ -11,6 +11,7 @@ from backend.blocks.ai_shortform_video_block import (
|
||||
from backend.blocks.apollo.organization import SearchOrganizationsBlock
|
||||
from backend.blocks.apollo.people import SearchPeopleBlock
|
||||
from backend.blocks.apollo.person import GetPersonDetailBlock
|
||||
from backend.blocks.codex import CodeGenerationBlock, CodexModel
|
||||
from backend.blocks.enrichlayer.linkedin import (
|
||||
GetLinkedinProfileBlock,
|
||||
GetLinkedinProfilePictureBlock,
|
||||
@@ -63,9 +64,10 @@ MODEL_COST: dict[LlmModel, int] = {
|
||||
LlmModel.O1_MINI: 4,
|
||||
# GPT-5 models
|
||||
LlmModel.GPT5: 2,
|
||||
LlmModel.GPT5_1: 5,
|
||||
LlmModel.GPT5_MINI: 1,
|
||||
LlmModel.GPT5_NANO: 1,
|
||||
LlmModel.GPT5_CHAT: 2,
|
||||
LlmModel.GPT5_CHAT: 5,
|
||||
LlmModel.GPT41: 2,
|
||||
LlmModel.GPT41_MINI: 1,
|
||||
LlmModel.GPT4O_MINI: 1,
|
||||
@@ -265,6 +267,20 @@ BLOCK_COSTS: dict[Type[Block], list[BlockCost]] = {
|
||||
AIStructuredResponseGeneratorBlock: LLM_COST,
|
||||
AITextSummarizerBlock: LLM_COST,
|
||||
AIListGeneratorBlock: LLM_COST,
|
||||
CodeGenerationBlock: [
|
||||
BlockCost(
|
||||
cost_type=BlockCostType.RUN,
|
||||
cost_filter={
|
||||
"model": CodexModel.GPT5_1_CODEX,
|
||||
"credentials": {
|
||||
"id": openai_credentials.id,
|
||||
"provider": openai_credentials.provider,
|
||||
"type": openai_credentials.type,
|
||||
},
|
||||
},
|
||||
cost_amount=5,
|
||||
)
|
||||
],
|
||||
CreateTalkingAvatarVideoBlock: [
|
||||
BlockCost(
|
||||
cost_amount=15,
|
||||
|
||||
Reference in New Issue
Block a user