flock-core 0.3.10__tar.gz → 0.3.13__tar.gz

{flock_core-0.3.10 → flock_core-0.3.13}/.gitignore

@@ -28,7 +28,9 @@ MANIFEST
 logs/
 metrics/
 .claude/
-
+concept_memory/
+news/
+.chroma/
 
 # PyInstaller
 #  Usually these files are written by a python script from a template

{flock_core-0.3.10 → flock_core-0.3.13}/PKG-INFO

@@ -1,6 +1,6 @@
 Metadata-Version: 2.4
 Name: flock-core
-Version: 0.3.10
+Version: 0.3.13
 Summary: Declarative LLM Orchestration at Scale
 Author-email: Andre Ratzenberger <andre.ratzenberger@whiteduck.de>
 License-File: LICENSE
@@ -8,6 +8,7 @@ Classifier: License :: OSI Approved :: MIT License
 Classifier: Operating System :: OS Independent
 Classifier: Programming Language :: Python :: 3
 Requires-Python: >=3.10
+Requires-Dist: chromadb>=0.6.3
 Requires-Dist: cloudpickle>=3.1.1
 Requires-Dist: devtools>=0.12.2
 Requires-Dist: dspy==2.5.42
@@ -30,6 +31,7 @@ Requires-Dist: python-box>=7.3.2
 Requires-Dist: python-decouple>=3.8
 Requires-Dist: questionary>=2.1.0
 Requires-Dist: rich>=13.9.4
+Requires-Dist: sentence-transformers>=3.4.1
 Requires-Dist: temporalio>=1.9.0
 Requires-Dist: tiktoken>=0.8.0
 Requires-Dist: toml>=0.10.2
@@ -39,7 +41,6 @@ Requires-Dist: zep-python>=2.0.2
 Provides-Extra: all
 Requires-Dist: docling>=2.18.0; extra == 'all'
 Requires-Dist: markdownify>=0.14.1; extra == 'all'
-Requires-Dist: sentence-transformers>=3.4.1; extra == 'all'
 Requires-Dist: tavily-python>=0.5.0; extra == 'all'
 Provides-Extra: tools
 Requires-Dist: markdownify>=0.14.1; extra == 'tools'

{flock_core-0.3.10 → flock_core-0.3.13}/examples/01_introduction/01_simple_example.py

@@ -1,6 +1,5 @@
 import os
-from flock.core import Flock, FlockFactory
-from flock.core.flock_api import FlockAPI
+from flock.core import Flock, FlockFactory 
 
 
 # --------------------------------
@@ -9,9 +8,7 @@ from flock.core.flock_api import FlockAPI
 # Flock uses litellm to talk to LLMs
 # Please consult the litellm documentation for valid IDs:
 # https://docs.litellm.ai/docs/providers
-
-
-MODEL = "azure/ara-gpt4o"
+MODEL = "openai/gpt-4o"
 
 
 # --------------------------------
@@ -41,6 +38,6 @@ flock.add_agent(bloggy)
 # Tell the flock who is the starting and what input to give it
 flock.run(
     start_agent=bloggy, 
-    input={"blog_idea": "A blog about cats"}
+    input={"blog_idea": "A blog about robot kittens"}
 )
 

{flock_core-0.3.10 → flock_core-0.3.13}/examples/03_misc/api/api_server.py

@@ -15,6 +15,5 @@ flock.add_agent(bloggy)
 
 # Swagger: http://127.0.0.1:8344/docs
 # Redoc: http://127.0.0.1:8344/redoc
-# POST: http://127.0.0.1:8344/run 
-api = FlockAPI(flock)
-api.start()
+# POST: http://127.0.0.1:8344/run
+flock.start_api()

flock_core-0.3.13/examples/03_misc/evaluator/multi_hops_memory.py

@@ -0,0 +1,80 @@
+
+from flock.core import Flock, FlockFactory
+from flock.core.logging.formatters.themes import OutputTheme
+from flock.evaluators.memory.memory_evaluator import MemoryEvaluator, MemoryEvaluatorConfig
+from flock.modules.memory.memory_module import MemoryModule, MemoryModuleConfig
+
+
+
+def write_to_kg():
+  write_to_kg_agent = FlockFactory.create_default_agent(model="openai/gpt-4o",name="write_to_kg_agent", 
+                                            input="data", 
+                                            output_theme=OutputTheme.aardvark_blue)
+  
+
+  write_to_kg_agent.evaluator = MemoryEvaluator(name="mem_eval", 
+                                config=MemoryEvaluatorConfig(splitting_mode="characters", 
+                                                          number_of_concepts_to_extract=3))
+
+  write_to_kg_agent.run(inputs={"data": "Andre is 38 years old and author of the agent framework 'flock'"})
+  write_to_kg_agent.run(inputs={"data": "Andre works for white duck"})
+  write_to_kg_agent.run(inputs={"data": "Andre has two cats"})
+  write_to_kg_agent.run(inputs={"data": "White Duck is a cloud consulting company"})
+  write_to_kg_agent.run(inputs={"data": "Flock is an agent framework designed for scalable multi-agent systems"})
+  write_to_kg_agent.run(inputs={"data": "One of Andre's cats is named Luna"})
+  write_to_kg_agent.run(inputs={"data": "The other cat is named Lucy"})
+  write_to_kg_agent.run(inputs={"data": "Andre lives in Germany"})
+  write_to_kg_agent.run(inputs={"data": "Germany is in Europe"})
+
+def read_from_kg():
+  read_from_kg_agent = FlockFactory.create_default_agent(model="openai/gpt-4o",name="read_from_kg_agent", 
+                                            input="query", 
+                                            output_theme=OutputTheme.aardvark_blue)
+
+  # replace the default evaluator with ZepEvaluator
+  read_from_kg_agent.evaluator = MemoryEvaluator(name="mem_eval", config=MemoryEvaluatorConfig())
+
+
+  # #### **2-Hop Question:**
+  # **Question:** What kind of company does the employer of the author of 'flock' belong to?  
+  # **Reasoning:**  
+  # 1. "Andre is the author of the agent framework 'flock'."  
+  # 2. "Andre works for White Duck."  
+  # 3. "White Duck is a cloud consulting company."  
+  # 4. Therefore, the employer of the author of 'flock' is a cloud consulting company.
+
+  read_from_kg_agent.run(inputs={"query": "What kind of company does the employer of the author of 'flock' belong to?"})
+
+def read_from_kg_and_evaluate():
+  read_from_kg_and_evaluate_agent = FlockFactory.create_default_agent(model="openai/gpt-4o",name="read_from_kg_and_evaluate_agent", 
+                                            input="query", 
+                                            output="short_answer",
+                                            output_theme=OutputTheme.aardvark_blue)
+  
+  read_from_kg_and_evaluate_agent.add_module(MemoryModule(name="mem_eval", config=MemoryModuleConfig(splitting_mode="characters")))
+
+
+
+  read_from_kg_and_evaluate_agent.run(inputs={"query": "What kind of company does the employer of the author of 'flock' belong to?"})
+
+  # #### **3-Hop Question:**
+  # **Question:** In which continent does the creator of the agent framework 'flock' live?  
+  # **Reasoning:**  
+  # 1. "Andre is the author of the agent framework 'flock'."  
+  # 2. "Andre lives in Germany."  
+  # 3. "Germany is in Europe."  
+  # 4. Therefore, the creator of 'flock' lives in Europe.
+  read_from_kg_and_evaluate_agent.run(inputs={"query": "In which continent does the creator of the agent framework 'flock' live? "})
+
+
+
+if __name__ == "__main__":
+  write_to_kg()
+ #read_from_kg()
+  #read_from_kg_and_evaluate()
+  pass
+
+
+
+
+

{flock_core-0.3.10 → flock_core-0.3.13}/examples/03_misc/modules/use_modules.py

@@ -25,6 +25,7 @@ from pprint import pprint
 from flock.core import Flock, FlockFactory
 from flock.core.logging.formatters.themes import OutputTheme
 from flock.core.tools import basic_tools
+from flock.modules.memory.memory_module import MemoryModule, MemoryModuleConfig
 from flock.modules.zep.zep_module import ZepModule, ZepModuleConfig
 
 
@@ -48,7 +49,11 @@ agent = FlockFactory.create_default_agent(
 # Modules run at certain points in the agent's lifecycle and can manipulate the inputs and outputs and the agent itself.
 # In this case, we're adding the Zep module to the agent, 
 # which allows it to use Zep to store and retrieve information in Knowledge Graphs.
-zep = ZepModule(name="zep",config=ZepModuleConfig())
+# Currently there are two graph based modules: Zep and Memory.
+# Memory is more lightweight and easier to use, but Zep offers more features and is more powerful.
+
+# zep = ZepModule(name="zep",config=ZepModuleConfig())
+zep = MemoryModule(name="mem_split",config=MemoryModuleConfig(splitting_mode="characters"))
 agent.add_module(zep)
 
 

{flock_core-0.3.10 → flock_core-0.3.13}/examples/03_misc/router/router_example.py

@@ -42,7 +42,6 @@ async def main():
     from flock.routers.llm.llm_router import LLMRouter, LLMRouterConfig
     
     llm_router = LLMRouter(
-        registry=None,  # Will be set by the framework
         config=LLMRouterConfig(
             temperature=0.1,  # Lower temperature for more deterministic decisions
             confidence_threshold=0.7,  # Higher threshold for more confident decisions

flock_core-0.3.13/examples/notebooks_wip/rag_comparison/rag.ipynb

@@ -0,0 +1,283 @@
+{
+ "cells": [
+  {
+   "cell_type": "markdown",
+   "metadata": {},
+   "source": [
+    "# RAG Comparison"
+   ]
+  },
+  {
+   "cell_type": "markdown",
+   "metadata": {},
+   "source": [
+    "### Import"
+   ]
+  },
+  {
+   "cell_type": "code",
+   "execution_count": 1,
+   "metadata": {},
+   "outputs": [],
+   "source": [
+    "# Imports\n",
+    "\n",
+    "from docling.document_converter import DocumentConverter\n",
+    "from tqdm import tqdm\n",
+    "from flock.core.tools.basic_tools import extract_links_from_markdown, get_web_content_as_markdown\n",
+    "# load all markdown file in news folder\n",
+    "import glob\n",
+    "import chromadb\n",
+    "from devtools import debug, pprint\n",
+    "from flock.core.tools.llm_tools import chunk_text_for_embedding\n",
+    "import random\n",
+    "from flock.core import FlockFactory\n",
+    "from flock.core.logging.formatters.themes import OutputTheme\n",
+    "from flock.evaluators.memory.memory_evaluator import MemoryEvaluator, MemoryEvaluatorConfig\n",
+    "from langchain_community.document_loaders import DirectoryLoader, TextLoader\n",
+    "from ragas.testset.graph import KnowledgeGraph\n",
+    "from ragas.testset.graph import Node, NodeType\n",
+    "from ragas.llms.base import llm_factory\n",
+    "from ragas.embeddings.base import embedding_factory\n",
+    "from ragas.testset.transforms import Parallel, apply_transforms\n",
+    "from ragas.testset.transforms import (\n",
+    "    HeadlinesExtractor,\n",
+    "    HeadlineSplitter,\n",
+    "    KeyphrasesExtractor,\n",
+    "    OverlapScoreBuilder,\n",
+    ")\n",
+    "from dataclasses import dataclass\n",
+    "import typing as t\n",
+    "from ragas.testset.synthesizers.multi_hop.base import (\n",
+    "    MultiHopQuerySynthesizer,\n",
+    "    MultiHopScenario,\n",
+    ")\n",
+    "from IPython.display import clear_output\n",
+    "\n"
+   ]
+  },
+  {
+   "cell_type": "markdown",
+   "metadata": {},
+   "source": [
+    "### Download News"
+   ]
+  },
+  {
+   "cell_type": "code",
+   "execution_count": null,
+   "metadata": {},
+   "outputs": [],
+   "source": [
+    "# NEWS import\n",
+    "import os\n",
+    "\n",
+    "\n",
+    "path = \"https://lite.cnn.com\"\n",
+    "\n",
+    "index_markdown = get_web_content_as_markdown(path)\n",
+    "links = extract_links_from_markdown(index_markdown,path)\n",
+    "\n",
+    "# create news folder if it doesn't exist\n",
+    "if not os.path.exists(\"news\"):\n",
+    "    os.makedirs(\"news\") \n",
+    "\n",
+    "\n",
+    "converter = DocumentConverter()\n",
+    "for link in tqdm(links):\n",
+    "    try:    \n",
+    "        result = converter.convert(link)\n",
+    "        markdown = result.document.export_to_markdown()\n",
+    "        with open(f\"news/{link.split('/')[-2]}.md\", \"w\", encoding=\"utf-8\") as f:\n",
+    "            f.write(markdown)\n",
+    "    except Exception as e:\n",
+    "        print(f\"Error converting {link}: {e}\")\n"
+   ]
+  },
+  {
+   "cell_type": "markdown",
+   "metadata": {},
+   "source": [
+    "### ChromaDB"
+   ]
+  },
+  {
+   "cell_type": "markdown",
+   "metadata": {},
+   "source": [
+    "#### Generate Embeddings and Collection"
+   ]
+  },
+  {
+   "cell_type": "code",
+   "execution_count": null,
+   "metadata": {},
+   "outputs": [],
+   "source": [
+    "\n",
+    "news_files = glob.glob(\"news/*.md\")\n",
+    "news_files = [open(file, \"r\", encoding=\"utf-8\") for file in news_files]\n",
+    "news_files = [(file.read(),file.name) for file in news_files]\n",
+    "\n",
+    "news_files_chunks = []\n",
+    "for file_content, file_name in tqdm(news_files):\n",
+    "    news_files_chunks.extend(chunk_text_for_embedding(file_content, file_name.split(\"\\\\\")[-1], chunk_size=2000, overlap=250))\n",
+    "\n",
+    "clear_output(wait=True) \n",
+    "\n",
+    "chroma_client = chromadb.PersistentClient(path=\".chroma\")\n",
+    "collection = chroma_client.get_or_create_collection(name=\"my_cnn_news\")        \n",
+    "for chunk in tqdm(news_files_chunks):\n",
+    "    collection.add(\n",
+    "        ids=[chunk[\"chunk_id\"]],\n",
+    "        documents=[chunk[\"text\"]],\n",
+    "        metadatas=[{\"source\": chunk[\"file\"]}],\n",
+    "    )\n",
+    "    "
+   ]
+  },
+  {
+   "cell_type": "markdown",
+   "metadata": {},
+   "source": [
+    "#### Sample Query"
+   ]
+  },
+  {
+   "cell_type": "code",
+   "execution_count": null,
+   "metadata": {},
+   "outputs": [],
+   "source": [
+    "chroma_client = chromadb.PersistentClient(path=\".chroma\")\n",
+    "collection = chroma_client.get_or_create_collection(name=\"my_cnn_news\")        \n",
+    "\n",
+    "query = \"What recent actions have been taken by the Trump administration regarding the federal workforce, and how does this relate to the broader context of Trump's policies and proposals?\"\n",
+    "results = collection.query(query_texts=[query], n_results=3)\n",
+    "debug(results)\n",
+    "\n"
+   ]
+  },
+  {
+   "cell_type": "markdown",
+   "metadata": {},
+   "source": [
+    "#### Generate Queries with Flock"
+   ]
+  },
+  {
+   "cell_type": "markdown",
+   "metadata": {},
+   "source": [
+    "Chunk Query"
+   ]
+  },
+  {
+   "cell_type": "code",
+   "execution_count": null,
+   "metadata": {},
+   "outputs": [],
+   "source": [
+    "chroma_client = chromadb.PersistentClient(path=\".chroma\")\n",
+    "collection = chroma_client.get_or_create_collection(name=\"my_cnn_news\")\n",
+    "entries = collection.get()\n",
+    "# Get 10 random entries from the collection\n",
+    "random_ids = random.sample(entries.get(\"ids\"), 10)\n",
+    "random_news_chunks = collection.get(ids=random_ids).get(\"documents\")\n",
+    "\n",
+    "# Easiest form of query - answer a question based on the chunk\n",
+    "simple_qa_generator = FlockFactory.create_default_agent(name=\"simple_qa_generator\",\n",
+    "                                                     model=\"openai/gpt-4o\",\n",
+    "                                                     input=\"rag_context: str\",\n",
+    "                                                     output=\"factoid_question: str, factoid_answer: str\",\n",
+    "                                                     enable_rich_tables=True)\n",
+    "\n",
+    "for news_chunk in random_news_chunks: \n",
+    "    qa_pair = await simple_qa_generator.run_async(inputs={\"rag_context\": news_chunk})\n",
+    "    print(qa_pair)\n",
+    "    \n",
+    "\n",
+    "\n",
+    "\n",
+    "\n"
+   ]
+  },
+  {
+   "cell_type": "markdown",
+   "metadata": {},
+   "source": [
+    "Document Query"
+   ]
+  },
+  {
+   "cell_type": "code",
+   "execution_count": null,
+   "metadata": {},
+   "outputs": [],
+   "source": [
+    "# More complex form of query - answer a question based on the whole document\n",
+    "# get 10 random news markdown files from the news folder\n",
+    "\n",
+    "news_files = glob.glob(\"news/*.md\")\n",
+    "news_files = [open(file, \"r\", encoding=\"utf-8\") for file in news_files]\n",
+    "news_files = [(file.read(),file.name) for file in news_files]\n",
+    "random_news_files = random.sample(news_files, 10)\n",
+    "\n",
+    "simple_qa_generator = FlockFactory.create_default_agent(name=\"simple_qa_generator\",\n",
+    "                                                     model=\"openai/gpt-4o\",\n",
+    "                                                     input=\"rag_context: str\",\n",
+    "                                                     output=\"factoid_question: str, factoid_answer: str, ground_truth: str\",\n",
+    "                                                     enable_rich_tables=True)\n",
+    "\n",
+    "# let the agent iterate over the news files and generate a question and answer for each file    \n",
+    "for news_file in random_news_files: \n",
+    "    qa_pair = await simple_qa_generator.run_async(inputs={\"rag_context\": news_file})\n",
+    "    print(qa_pair)"
+   ]
+  },
+  {
+   "cell_type": "code",
+   "execution_count": null,
+   "metadata": {},
+   "outputs": [],
+   "source": [
+    "\n",
+    "\n",
+    "\n",
+    "write_to_kg_agent = FlockFactory.create_default_agent(model=\"openai/gpt-4o\",name=\"write_to_kg_agent\", \n",
+    "                                            input=\"data\", \n",
+    "                                            output_theme=OutputTheme.aardvark_blue)\n",
+    "\n",
+    "\n",
+    "write_to_kg_agent.evaluator = MemoryEvaluator(name=\"mem_eval\", \n",
+    "                                              config=MemoryEvaluatorConfig(splitting_mode=\"none\", \n",
+    "                                                                           number_of_concepts_to_extract=5))\n",
+    "\n",
+    "\n",
+    "\n",
+    "\n"
+   ]
+  }
+ ],
+ "metadata": {
+  "kernelspec": {
+   "display_name": ".venv",
+   "language": "python",
+   "name": "python3"
+  },
+  "language_info": {
+   "codemirror_mode": {
+    "name": "ipython",
+    "version": 3
+   },
+   "file_extension": ".py",
+   "mimetype": "text/x-python",
+   "name": "python",
+   "nbconvert_exporter": "python",
+   "pygments_lexer": "ipython3",
+   "version": "3.12.8"
+  }
+ },
+ "nbformat": 4,
+ "nbformat_minor": 2
+}