Wednesday, May 13, 2026

CrewAI "Hello World" in a single colab sheet, without YAML files

 [Cell 001]

!pip install -U crewai


[Cell 002]
from google.colab import userdata
import os
os.environ["GOOGLE_API_KEY"] = userdata.get('GEMINI_API_KEY_006')
os.environ["MODEL"] = "gemini/gemini-3.1-flash-lite"

#EVEN THE BELOW WILL ALSO WORK
# from google.colab import userdata

# # This sets it for both Python AND any !shell commands you run later
# %env GEMINI_API_KEY={userdata.get('GEMINI_API_KEY_006')}
## %env MODEL=gemini/gemini-3.1-flash-lite #if model is required  in env

[Cell 003]
#!/usr/bin/env python
from crewai import Agent, Crew, Process, Task
from crewai.project import CrewBase, agent, crew, task
from crewai.agents.agent_builder.base_agent import BaseAgent
# If you want to run a snippet of code before or after the crew starts,
# you can use the @before_kickoff and @after_kickoff decorators
# https://docs.crewai.com/concepts/crews#example-crew-class-with-decorators

@CrewBase
class HelloWorld():
    """HelloWorld crew"""

    agents: list[BaseAgent]
    tasks: list[Task]

    # Learn more about YAML configuration files here:
    # Agents: https://docs.crewai.com/concepts/agents#yaml-configuration-recommended
    # Tasks: https://docs.crewai.com/concepts/tasks#yaml-configuration-recommended
   
    # If you would like to add tools to your agents, you can learn more about it here:
    # https://docs.crewai.com/concepts/agents#agent-tools
    @agent
    def researcher(self) -> Agent:
        return Agent(
            #config=self.agents_config['researcher'], # type: ignore[index]
            role="{topic} Senior Data Researcher",
            goal="Uncover cutting-edge developments in {topic}",
            backstory="""You're a seasoned researcher with a knack for uncovering the latest
            developments in {topic}. Known for your ability to find the most relevant
            information and present it in a clear and concise manner.""",
            verbose=True
        )

    @agent
    def reporting_analyst(self) -> Agent:
        return Agent(
            #config=self.agents_config['reporting_analyst'], # type: ignore[index]
            role="{topic} Reporting Analyst",
            goal="Create detailed reports based on {topic} data analysis and research findings",
            backstory="""You're a meticulous analyst with a keen eye for detail. You're known for
            your ability to turn complex data into clear and concise reports, making
            it easy for others to understand and act on the information you provide.""",
            verbose=True
        )

    # To learn more about structured task outputs,
    # task dependencies, and task callbacks, check out the documentation:
    # https://docs.crewai.com/concepts/tasks#overview-of-a-task
    @task
    def research_task(self) -> Task:
        return Task(
            #config=self.tasks_config['research_task'], # type: ignore[index]
            description="""Conduct a thorough research about {topic}
            Make sure you find any interesting and relevant information given
            the current year is {current_year}.
            """,
            expected_output="""A list with 10 bullet points of the most relevant information about {topic}""",
            #agent="researcher"
            agent=self.researcher()            
        )

    @task
    def reporting_task(self) -> Task:
        return Task(
            #config=self.tasks_config['reporting_task'], # type: ignore[index]
            description="""Review the context you got and expand each topic into a full section for a report.
            Make sure the report is detailed and contains any and all relevant information.""",
            expected_output="""A fully fledged report with the main topics, each with a full section of information.
            Formatted as markdown without '```'""",
            #agent="reporting_analyst",
            agent=self.reporting_analyst(),
            output_file='report.md'
        )

    @crew
    def crew(self) -> Crew:
        """Creates the HelloWorld crew"""
        # To learn how to add knowledge sources to your crew, check out the documentation:
        # https://docs.crewai.com/concepts/knowledge#what-is-knowledge

        return Crew(
            #agents=self.agents, # Automatically created by the @agent decorator
            #tasks=self.tasks, # Automatically created by the @task decorator
            agents=[self.researcher(), self.reporting_analyst()],
            tasks=[self.research_task(), self.reporting_task()],
            process=Process.sequential,
            verbose=True
            # process=Process.hierarchical, # In case you wanna use that instead https://docs.crewai.com/how-to/Hierarchical/
        )


import sys
import warnings

from datetime import datetime

#from hello_world.crew import HelloWorld

warnings.filterwarnings("ignore", category=SyntaxWarning, module="pysbd")

# This main file is intended to be a way for you to run your
# crew locally, so refrain from adding unnecessary logic into this file.
# Replace with inputs you want to test with, it will automatically
# interpolate any tasks and agents information

def run():
    """
    Run the crew.
    """
    inputs = {
        'topic': 'AI LLMs',
        'current_year': str(datetime.now().year)
    }
    #print(inputs)
    #print(type(inputs))  
    try:
        #HelloWorld().crew().kickoff(inputs={"topic" : "AI LLMs", "current_year" : 2026})
        HelloWorld().crew().kickoff(inputs=inputs)
    except Exception as e:
        raise Exception(f"An error occurred while running the crew: {e}")


def train():
    """
    Train the crew for a given number of iterations.
    """
    inputs = {
        "topic": "AI LLMs",
        'current_year': str(datetime.now().year)
    }
    try:
        HelloWorld().crew().train(n_iterations=int(sys.argv[1]), filename=sys.argv[2], inputs=inputs)

    except Exception as e:
        raise Exception(f"An error occurred while training the crew: {e}")

def replay():
    """
    Replay the crew execution from a specific task.
    """
    try:
        HelloWorld().crew().replay(task_id=sys.argv[1])

    except Exception as e:
        raise Exception(f"An error occurred while replaying the crew: {e}")

def test():
    """
    Test the crew execution and returns the results.
    """
    inputs = {
        "topic": "AI LLMs",
        "current_year": str(datetime.now().year)
    }

    try:
        HelloWorld().crew().test(n_iterations=int(sys.argv[1]), eval_llm=sys.argv[2], inputs=inputs)

    except Exception as e:
        raise Exception(f"An error occurred while testing the crew: {e}")

def run_with_trigger():
    """
    Run the crew with trigger payload.
    """
    import json

    if len(sys.argv) < 2:
        raise Exception("No trigger payload provided. Please provide JSON payload as argument.")

    try:
        trigger_payload = json.loads(sys.argv[1])
    except json.JSONDecodeError:
        raise Exception("Invalid JSON payload provided as argument")

    inputs = {
        "crewai_trigger_payload": trigger_payload,
        "topic": "",
        "current_year": ""
    }

    try:
        result = HelloWorld().crew().kickoff(inputs=inputs)
        return result
    except Exception as e:
        raise Exception(f"An error occurred while running the crew with trigger: {e}")

run()



====================================================================================
OUTPUT
====================================================================================
WARNING:root:File not found: /content/config/agents.yaml
WARNING:root:Agent config file not found at /content/config/agents.yaml. Proceeding with empty agent configurations.
WARNING:root:File not found: /content/config/tasks.yaml
WARNING:root:Task config file not found at /content/config/tasks.yaml. Proceeding with empty task configurations.
{'topic': 'AI LLMs', 'current_year': 2026}
<class 'dict'>
╭─────────────────────────────────────────── 🚀 Crew Execution Started ───────────────────────────────────────────╮
│                                                                                                                 │
│  Crew Execution Started                                                                                         │
│  Name: HelloWorld                                                                                               │
│  ID: 679ff0a1-5264-4ef8-97b6-a3fd8b9935e5                                                                       │
│                                                                                                                 │
│                                                                                                                 │
╰─────────────────────────────────────────────────────────────────────────────────────────────────────────────────╯

╭──────────────────────────────────────────────── 📋 Task Started ────────────────────────────────────────────────╮
│                                                                                                                 │
│  Task Started                                                                                                   │
│  Name: research_task                                                                                            │
│  ID: 975f337f-37e0-4b94-9db2-5e016ac68cf0                                                                       │
│                                                                                                                 │
│                                                                                                                 │
╰─────────────────────────────────────────────────────────────────────────────────────────────────────────────────╯

╭─────────────────────────────────────────────── 🤖 Agent Started ────────────────────────────────────────────────╮
│                                                                                                                 │
│  Agent: AI LLMs Senior Data Researcher                                                                          │
│                                                                                                                 │
│  Task: Conduct a thorough research about AI LLMs                                                                │
│              Make sure you find any interesting and relevant information given                                  │
│              the current year is 2026.                                                                          │
│                                                                                                                 │
│                                                                                                                 │
╰─────────────────────────────────────────────────────────────────────────────────────────────────────────────────╯

╭───────────────────────────────────────────── ✅ Agent Final Answer ─────────────────────────────────────────────╮
│                                                                                                                 │
│  Agent: AI LLMs Senior Data Researcher                                                                          │
│                                                                                                                 │
│  Final Answer:                                                                                                  │
│  As of mid-2026, the landscape of Large Language Models has shifted from "scale-at-all-costs" to "efficiency,   │
│  reasoning, and multimodal integration." The focus is no longer just on parameter count, but on the capability  │
│  of models to act as autonomous agents capable of long-horizon planning.                                        │
│                                                                                                                 │
│  Here are the 10 most relevant developments in AI LLMs for 2026:                                                │
│                                                                                                                 │
│  *   **The Rise of Inference-Time Compute:** Models are no longer limited to the "system 1" (fast, intuitive)   │
│  thinking of 2024. We have seen a massive industry shift toward "System 2" reasoning, where models leverage     │
│  significant compute during the *inference* phase to verify their own steps, perform tree-of-thought searches,  │
│  and self-correct before producing a final token.                                                               │
│  *   **Massive Context Windows are Now Standard:** We have moved beyond the "lost in the middle" phenomenon.    │
│  By 2026, models routinely handle 10-million+ token context windows with near-perfect retrieval accuracy,       │
│  allowing for the ingestion of entire corporate codebases or multi-year legal archives into a single prompt.    │
│  *   **Neural-Symbolic Hybridization:** To solve the "hallucination" problem that plagued LLMs in the           │
│  mid-2020s, current state-of-the-art models now integrate symbolic logic engines. These models act as           │
│  orchestrators, offloading mathematical and logical proofs to deterministic solvers while using the LLM for     │
│  natural language interpretation.                                                                               │
│  *   **Native Multimodality:** We have abandoned the "model-stitching" approach (where a vision encoder was     │
│  patched onto a text decoder). Today’s SOTA models are natively trained on interleaved video, audio, and text,  │
│  allowing for "real-time" reasoning where the AI can watch a live video feed and debug code or analyze          │
│  physical world processes in milliseconds.                                                                      │
│  *   **Agentic Orchestration Frameworks:** The focus has shifted from "chatbots" to "agents." Most enterprise   │
│  LLMs in 2026 exist within sophisticated frameworks that allow them to autonomously navigate the web, execute   │
│  shell commands, and interact with private APIs to complete complex, multi-step business goals without human    │
│  intervention.                                                                                                  │
│  *   **The Energy-Efficiency Revolution:** Following the massive power consumption spikes of 2024-2025, there   │
│  has been a breakthrough in "Small Language Models" (SLMs). Distillation techniques have reached a point where  │
│  models with performance parity to 2024’s largest models can run locally on mobile devices and edge hardware    │
│  with minimal thermal output.                                                                                   │
│  *   **Synthetic Data Pipelines:** Given that we have essentially exhausted the "high-quality" human-written    │
│  text on the public internet, models are now being trained predominantly on "model-generated data." The         │
│  industry has developed robust "automated curriculum learning" where a teacher model creates and filters        │
│  educational datasets for the student model to learn from.                                                      │
│  *   **Privacy-Preserving Federated Training:** Enterprises now utilize federated learning techniques that      │
│  allow LLMs to be fine-tuned on sensitive internal data without the data ever leaving the client's local        │
│  servers. This has unlocked the adoption of LLMs in highly regulated industries like healthcare and defense.    │
│  *   **Long-Term Memory Persistence:** We have transitioned from ephemeral chat sessions to "persistent         │
│  persona memory." LLMs now maintain a vector-based episodic memory store that evolves over time, allowing the   │
│  model to recall specific interactions from months prior, creating a more personalized and consistent user      │
│  experience.                                                                                                    │
│  *   **Regulatory "Watermarking" as a Standard:** Due to the 2025 Global AI Safety Accords, all major           │
│  commercial models now feature mandatory, cryptographically secure watermarking at the token level. This is     │
│  now natively supported at the hardware level in modern server GPUs to ensure provenance and combat the         │
│  proliferation of deepfakes and AI-generated disinformation.                                                    │
│                                                                                                                 │
╰─────────────────────────────────────────────────────────────────────────────────────────────────────────────────╯

╭────────────────────────────────────────────── 📋 Task Completion ───────────────────────────────────────────────╮
│                                                                                                                 │
│  Task Completed                                                                                                 │
│  Name: research_task                                                                                            │
│  Agent: AI LLMs Senior Data Researcher                                                                          │
│                                                                                                                 │
│                                                                                                                 │
╰─────────────────────────────────────────────────────────────────────────────────────────────────────────────────╯

╭──────────────────────────────────────────────── 📋 Task Started ────────────────────────────────────────────────╮
│                                                                                                                 │
│  Task Started                                                                                                   │
│  Name: reporting_task                                                                                           │
│  ID: 67f7a202-157f-4429-ab7c-9589866b7b82                                                                       │
│                                                                                                                 │
│                                                                                                                 │
╰─────────────────────────────────────────────────────────────────────────────────────────────────────────────────╯

╭─────────────────────────────────────────────── 🤖 Agent Started ────────────────────────────────────────────────╮
│                                                                                                                 │
│  Agent: AI LLMs Reporting Analyst                                                                               │
│                                                                                                                 │
│  Task: Review the context you got and expand each topic into a full section for a report.                       │
│              Make sure the report is detailed and contains any and all relevant information.                    │
│                                                                                                                 │
╰─────────────────────────────────────────────────────────────────────────────────────────────────────────────────╯

╭───────────────────────────────────────────── ✅ Agent Final Answer ─────────────────────────────────────────────╮
│                                                                                                                 │
│  Agent: AI LLMs Reporting Analyst                                                                               │
│                                                                                                                 │
│  Final Answer:                                                                                                  │
│  # AI LLM Landscape Report: Mid-2026 Industry Analysis                                                          │
│                                                                                                                 │
│  ## Introduction                                                                                                │
│  As of mid-2026, the artificial intelligence landscape has undergone a paradigm shift. The industry has moved   │
│  away from the "bigger is better" methodology that defined the early 2020s, favoring a strategy built on        │
│  operational efficiency, cognitive reasoning, and deep multimodal integration. The focus has transitioned from  │
│  passive conversational assistants to autonomous agents capable of complex, long-horizon planning and           │
│  real-world task execution. This report outlines the ten critical technological and structural advancements     │
│  currently shaping the AI sector.                                                                               │
│                                                                                                                 │
│  ## 1. The Rise of Inference-Time Compute                                                                       │
│  The limitations of "System 1" (fast, intuitive) thinking—where a model simply predicts the next token—have     │
│  been surpassed by "System 2" reasoning capabilities. In 2026, state-of-the-art models are designed to          │
│  allocate substantial compute resources during the inference phase rather than just the training phase. By      │
│  utilizing techniques such as tree-of-thought searches, self-verification cycles, and iterative                 │
│  self-correction, models now validate their logic before finalizing an output. This shift has drastically       │
│  reduced logic errors and increased the reliability of complex decision-making processes.                       │
│                                                                                                                 │
│  ## 2. Massive Context Windows as Standard                                                                      │
│  The "lost in the middle" retrieval phenomenon has been effectively mitigated. Modern LLMs now treat            │
│  10-million+ token context windows as the industry baseline. This capability allows users to upload entire      │
│  corporate codebases, years of legal archives, or comprehensive technical documentation into a single prompt.   │
│  Because retrieval accuracy is now near-perfect, these models serve as a "unified memory" for enterprise        │
│  operations, enabling deep analysis across massive datasets without the need for cumbersome external RAG        │
│  (Retrieval-Augmented Generation) architectures.                                                                │
│                                                                                                                 │
│  ## 3. Neural-Symbolic Hybridization                                                                            │
│  To address the pervasive issue of hallucinations, the industry has adopted neural-symbolic hybridization. By   │
│  integrating symbolic logic engines with neural networks, AI models now serve as orchestrators. When a task     │
│  requires rigid mathematical precision or logical verification, the LLM offloads the work to deterministic,     │
│  rule-based solvers. The LLM remains responsible for natural language processing and task orchestration, while  │
│  the symbolic engine ensures the integrity of the data, resulting in a hybrid system that is both flexible and  │
│  mathematically infallible.                                                                                     │
│                                                                                                                 │
│  ## 4. Native Multimodality                                                                                     │
│  The era of "model-stitching"—patching vision encoders onto text decoders—is over. SOTA models in 2026 are      │
│  natively trained on interleaved streams of video, audio, and text from the outset. This holistic architecture  │
│  allows for real-time, low-latency reasoning. For example, a model can process a live video feed of an          │
│  industrial robotic process, identify a mechanical failure, and propose a code-level fix in milliseconds. This  │
│  native integration has transformed AI from a text-based tool into a sensory-aware agent capable of navigating  │
│  the physical world.                                                                                            │
│                                                                                                                 │
│  ## 5. Agentic Orchestration Frameworks                                                                         │
│  Chatbots have been relegated to simple query-response tools; the current standard is the "Agentic Framework."  │
│  Modern LLMs are integrated into environments that allow them to autonomously navigate the web, execute shell   │
│  commands, manage file systems, and interface with private APIs. These frameworks permit agents to complete     │
│  complex, multi-step business objectives—such as drafting a contract, deploying the associated code, and        │
│  filing a regulatory report—with only high-level oversight from human operators.                                │
│                                                                                                                 │
│  ## 6. The Energy-Efficiency Revolution                                                                         │
│  Following the power infrastructure strain of 2024-2025, the industry successfully pivoted toward extreme       │
│  optimization. Breakthroughs in model distillation have resulted in "Small Language Models" (SLMs) that match   │
│  the performance of legacy giant models while operating with a fraction of the thermal and energy footprint.    │
│  These efficient models now run locally on mobile devices and edge hardware, enabling high-performance,         │
│  private, and offline AI capabilities that were previously impossible due to energy constraints.                │
│                                                                                                                 │
│  ## 7. Synthetic Data Pipelines                                                                                 │
│  Public internet data has been exhausted as a training resource, leading to the rise of robust synthetic data   │
│  pipelines. Models now learn primarily from "teacher-student" hierarchies where a highly sophisticated model    │
│  generates, validates, and filters educational datasets. This "automated curriculum learning" ensures that      │
│  training data is optimized for high-reasoning tasks, effectively creating a self-sustaining loop of            │
│  improvement that allows models to reach higher capabilities without relying on human-generated web content.    │
│                                                                                                                 │
│  ## 8. Privacy-Preserving Federated Training                                                                    │
│  In 2026, enterprise adoption in high-stakes fields like healthcare, defense, and finance has reached an        │
│  all-time high, driven by federated learning. This technique allows models to be fine-tuned on sensitive        │
│  internal data across distributed servers without the data ever being centralized or exposed. By keeping        │
│  training local, organizations maintain full compliance with data sovereignty laws, ensuring that the model     │
│  learns the organization's unique requirements without risking intellectual property or personal identifiable   │
│  information (PII).                                                                                             │
│                                                                                                                 │
│  ## 9. Long-Term Memory Persistence                                                                             │
│  The industry has abandoned ephemeral, session-based interactions. Modern LLMs incorporate vector-based         │
│  episodic memory stores that maintain a persistent "persona" or institutional history for the user. This        │
│  long-term memory allows the model to recall specific context, preferences, or tasks from interactions          │
│  occurring months or even years prior. This persistence facilitates a more consistent, personalized             │
│  experience, as the AI understands the user's long-term goals rather than treating each prompt as a blank       │
│  slate.                                                                                                         │
│                                                                                                                 │
│  ## 10. Regulatory "Watermarking" as a Standard                                                                 │
│  Following the 2025 Global AI Safety Accords, hardware-level support for provenance is now mandatory. All       │
│  major commercial LLMs now integrate cryptographically secure, token-level watermarking. This functionality is  │
│  baked into modern server-grade GPUs, ensuring that any content generated by an AI can be traced back to its    │
│  model and origin. This standardized system provides a critical line of defense against the proliferation of    │
│  deepfakes and disinformation, ensuring that human-generated and machine-generated content can be objectively   │
│  distinguished.                                                                                                 │
│                                                                                                                 │
╰─────────────────────────────────────────────────────────────────────────────────────────────────────────────────╯

╭────────────────────────────────────────────── 📋 Task Completion ───────────────────────────────────────────────╮
│                                                                                                                 │
│  Task Completed                                                                                                 │
│  Name: reporting_task                                                                                           │
│  Agent: AI LLMs Reporting Analyst                                                                               │
│                                                                                                                 │
│                                                                                                                 │
╰─────────────────────────────────────────────────────────────────────────────────────────────────────────────────╯

No comments:

Post a Comment

Comparing features of LangGraph vs CrewAI

LangGraph Orchestration vs CrewAI Orchestration The following table compares LangGraph and CrewAI orchestration across their menta...