parallelization_agent.py

import asyncio
import json

from strands import Agent, tool
from strands_tools import calculator
from strands.models import BedrockModel

bedrock_model = BedrockModel(
    model_id="us.anthropic.claude-3-7-sonnet-20250219-v1:0",
    region_name='us-east-1',
    temperature=0.3,
)


@tool
def write_linkedIn_post(content: str):
    '''

       :param content: {"title":"string","key_outcomes":["string"],"cta":"string"}

       :return:{"post":"string"}
       '''
    LINKEDIN_AGENT_SYSTEM_PROMPT = '''
    You write LinkedIn-ready announcements that are clear, professional, and skimmable.
    Success = a single post with a strong hook, value-focused bullets, and a clear CTA.
    Principles: no hashtags in the body, avoid jargon, keep it friendly-professional, 120–220 words max.
    '''

    linkedin_agent = Agent(
        model=bedrock_model,
        system_prompt=LINKEDIN_AGENT_SYSTEM_PROMPT,
        callback_handler=None
    )

    response = linkedin_agent(content)

    return response


@tool
def write_twitter_post(content: str):
    '''

    :param content: {"title":"string","key_outcomes":["string"],"cta":"string","max_tweets":5}

    :return:{"thread":["string","string","string","string","string"]}

    '''
    TWITTER_AGENT_SYSTEM_PROMPT = '''
    You craft concise, high-signal threads (≤5 tweets) that teach or announce.
    Success = a compelling first tweet, 3–4 crisp points, and a closing CTA.
    Principles: short sentences, plain language, minimal emojis, lightweight hashtags (≤2 total).
    '''

    twitter_agent = Agent(
        model=bedrock_model,
        system_prompt=TWITTER_AGENT_SYSTEM_PROMPT,
        callback_handler=None
    )

    response = twitter_agent(content)

    return response


@tool
def write_ig_post(content: str):
    """

       :param content:{"title":"string","vibe":"educational|inspirational","cta":"string"}

       :return:{"caption":"string","hashtags":["string"]}

       """
    IG_AGENT_SYSTEM_PROMPT = '''
   You write engaging IG captions with a warm, human tone.
    Success = 1–3 short lines, a simple CTA, and a tidy hashtag set.
    Principles: keep it friendly and inspirational, no walls of text, ≤10 relevant hashtags.
    '''

    ig_agent = Agent(
        model=bedrock_model,
        system_prompt=IG_AGENT_SYSTEM_PROMPT,
        callback_handler=None
    )

    response = ig_agent(content)

    return response


SYSTEM_PROMPT = '''
You are an aggregation agent that merges platform-specific drafts into a coherent promo pack.
Success = consistent voice and CTA across platforms, no duplication, cleaned formatting, and de-duplicated hashtags.
Principles: do not rewrite meaning; normalize tone, enforce platform rules (e.g., no hashtags in LinkedIn body), and return a tidy bundle labeled by channel.
'''
'''
{"bundle":{"linkedin":"string","x_thread":["string"],"instagram":{"caption":"string","hashtags":["string"]}},"notes":"string"}

'''
# Initialize our agent without a callback handler
aggregator_agent = Agent(
    model=bedrock_model,
    system_prompt=SYSTEM_PROMPT,
    tools=[write_linkedIn_post, write_ig_post, write_twitter_post],
    callback_handler=None
)

# Async function that iterates over streamed agent events

query = "What's the current state of AI in Africa wrt to growth areas. Include encouraging discussions and ideas, in an inspirational tone?"

agent_response = aggregator_agent(json.dumps(query))
print(agent_response)