From 920f931a21fa989ece4a0f9ab3d54a16fa019936 Mon Sep 17 00:00:00 2001
From: Toran Bruce Richards <toran.richards@gmail.com>
Date: Tue, 16 Jul 2024 11:51:37 +0100
Subject: [PATCH] feat(blocks): Add summariser block for recursive text
summarization functionality (#7431)
Co-authored-by: Zamil Majdy <zamil.majdy@agpt.co>
---
.../autogpt_server/blocks/{ai.py => llm.py} | 132 +++++++++++++++++-
.../autogpt_server/usecases/block_autogen.py | 2 +-
.../usecases/reddit_marketing.py | 2 +-
3 files changed, 128 insertions(+), 8 deletions(-)
rename rnd/autogpt_server/autogpt_server/blocks/{ai.py => llm.py} (59%)
diff --git a/rnd/autogpt_server/autogpt_server/blocks/ai.py b/rnd/autogpt_server/autogpt_server/blocks/llm.py
similarity index 59%
rename from rnd/autogpt_server/autogpt_server/blocks/ai.py
rename to rnd/autogpt_server/autogpt_server/blocks/llm.py
index 174bbf9c9..7f0cb9692 100644
--- a/rnd/autogpt_server/autogpt_server/blocks/ai.py
+++ b/rnd/autogpt_server/autogpt_server/blocks/llm.py
@@ -2,8 +2,8 @@ import logging
from enum import Enum
from typing import NamedTuple
-import openai
import anthropic
+import openai
from groq import Groq
from autogpt_server.data.block import Block, BlockOutput, BlockSchema, BlockFieldSecret
@@ -92,15 +92,21 @@ class LlmCallBlock(Block):
)
@staticmethod
- def llm_call(api_key: str, model: LlmModel, prompt: list[dict], json_format: bool) -> str:
+ def llm_call(
+ api_key: str,
+ model: LlmModel,
+ prompt: list[dict],
+ json_format: bool
+ ) -> str:
provider = model.metadata.provider
-
+
if provider == "openai":
openai.api_key = api_key
+ response_format = {"type": "json_object"} if json_format else None
response = openai.chat.completions.create(
model=model.value,
messages=prompt, # type: ignore
- response_format={"type": "json_object"} if json_format else None, # type: ignore
+ response_format=response_format, # type: ignore
)
return response.choices[0].message.content or ""
elif provider == "anthropic":
@@ -116,10 +122,11 @@ class LlmCallBlock(Block):
return response.content[0].text if response.content else ""
elif provider == "groq":
client = Groq(api_key=api_key)
+ response_format = {"type": "json_object"} if json_format else None
response = client.chat.completions.create(
model=model.value,
messages=prompt, # type: ignore
- response_format={"type": "json_object"} if json_format else None,
+ response_format=response_format, # type: ignore
)
return response.choices[0].message.content or ""
else:
@@ -136,7 +143,10 @@ class LlmCallBlock(Block):
prompt.append({"role": "system", "content": input_data.sys_prompt})
if input_data.expected_format:
- expected_format = [f'"{k}": "{v}"' for k, v in input_data.expected_format.items()]
+ expected_format = [
+ f'"{k}": "{v}"' for k, v in
+ input_data.expected_format.items()
+ ]
format_prompt = ",\n ".join(expected_format)
sys_prompt = trim_prompt(f"""
|Reply in json format:
@@ -199,3 +209,113 @@ class LlmCallBlock(Block):
retry_prompt = f"Error calling LLM: {e}"
yield "error", retry_prompt
+
+
+class TextSummarizerBlock(Block):
+ class Input(BlockSchema):
+ text: str
+ model: LlmModel = LlmModel.GPT4_TURBO
+ api_key: BlockFieldSecret = BlockFieldSecret(value="")
+ # TODO: Make this dynamic
+ max_tokens: int = 4000 # Adjust based on the model's context window
+ chunk_overlap: int = 100 # Overlap between chunks to maintain context
+
+ class Output(BlockSchema):
+ summary: str
+ error: str
+
+ def __init__(self):
+ super().__init__(
+ id="c3d4e5f6-7g8h-9i0j-1k2l-m3n4o5p6q7r8",
+ input_schema=TextSummarizerBlock.Input,
+ output_schema=TextSummarizerBlock.Output,
+ test_input={"text": "Lorem ipsum..." * 100},
+ test_output=("summary", "Final summary of a long text"),
+ test_mock={
+ "llm_call": lambda input_data:
+ {"final_summary": "Final summary of a long text"}
+ if "final_summary" in input_data.expected_format
+ else {"summary": "Summary of a chunk of text"}
+ }
+ )
+
+ def run(self, input_data: Input) -> BlockOutput:
+ try:
+ for output in self._run(input_data):
+ yield output
+ except Exception as e:
+ yield "error", str(e)
+
+ def _run(self, input_data: Input) -> BlockOutput:
+ chunks = self._split_text(
+ input_data.text,
+ input_data.max_tokens,
+ input_data.chunk_overlap
+ )
+ summaries = []
+
+ for chunk in chunks:
+ chunk_summary = self._summarize_chunk(chunk, input_data)
+ summaries.append(chunk_summary)
+
+ final_summary = self._combine_summaries(summaries, input_data)
+ yield "summary", final_summary
+
+ @staticmethod
+ def _split_text(text: str, max_tokens: int, overlap: int) -> list[str]:
+ words = text.split()
+ chunks = []
+ chunk_size = max_tokens - overlap
+
+ for i in range(0, len(words), chunk_size):
+ chunk = " ".join(words[i:i + max_tokens])
+ chunks.append(chunk)
+
+ return chunks
+
+ @staticmethod
+ def llm_call(input_data: LlmCallBlock.Input) -> dict[str, str]:
+ llm_block = LlmCallBlock()
+ for output_name, output_data in llm_block.run(input_data):
+ if output_name == "response":
+ return output_data
+ raise ValueError("Failed to get a response from the LLM.")
+
+ def _summarize_chunk(self, chunk: str, input_data: Input) -> str:
+ prompt = f"Summarize the following text concisely:\n\n{chunk}"
+
+ llm_response = self.llm_call(LlmCallBlock.Input(
+ prompt=prompt,
+ api_key=input_data.api_key,
+ model=input_data.model,
+ expected_format={"summary": "The summary of the given text."}
+ ))
+
+ return llm_response["summary"]
+
+ def _combine_summaries(self, summaries: list[str], input_data: Input) -> str:
+ combined_text = " ".join(summaries)
+
+ if len(combined_text.split()) <= input_data.max_tokens:
+ prompt = ("Provide a final, concise summary of the following summaries:\n\n"
+ + combined_text)
+
+ llm_response = self.llm_call(LlmCallBlock.Input(
+ prompt=prompt,
+ api_key=input_data.api_key,
+ model=input_data.model,
+ expected_format={
+ "final_summary": "The final summary of all provided summaries."
+ }
+ ))
+
+ return llm_response["final_summary"]
+ else:
+ # If combined summaries are still too long, recursively summarize
+ return self._run(TextSummarizerBlock.Input(
+ text=combined_text,
+ api_key=input_data.api_key,
+ model=input_data.model,
+ max_tokens=input_data.max_tokens,
+ chunk_overlap=input_data.chunk_overlap
+ )).send(None)[1] # Get the first yielded value
diff --git a/rnd/autogpt_server/autogpt_server/usecases/block_autogen.py b/rnd/autogpt_server/autogpt_server/usecases/block_autogen.py
index f00dc8322..1e3cd22fa 100644
--- a/rnd/autogpt_server/autogpt_server/usecases/block_autogen.py
+++ b/rnd/autogpt_server/autogpt_server/usecases/block_autogen.py
@@ -1,6 +1,6 @@
from pathlib import Path
-from autogpt_server.blocks.ai import LlmCallBlock
+from autogpt_server.blocks.llm import LlmCallBlock
from autogpt_server.blocks.basic import ValueBlock
from autogpt_server.blocks.block import BlockInstallationBlock
from autogpt_server.blocks.http import HttpRequestBlock
diff --git a/rnd/autogpt_server/autogpt_server/usecases/reddit_marketing.py b/rnd/autogpt_server/autogpt_server/usecases/reddit_marketing.py
index d0cc025bc..d9230bbdd 100644
--- a/rnd/autogpt_server/autogpt_server/usecases/reddit_marketing.py
+++ b/rnd/autogpt_server/autogpt_server/usecases/reddit_marketing.py
@@ -1,5 +1,5 @@
from autogpt_server.data.graph import Graph, Link, Node, create_graph
-from autogpt_server.blocks.ai import LlmCallBlock
+from autogpt_server.blocks.llm import LlmCallBlock
from autogpt_server.blocks.reddit import (
RedditGetPostsBlock,
RedditPostCommentBlock,