Upload app.py

app.py

@@ -0,0 +1,421 @@
+import os
+import textwrap
+
+# ─── CrewAI core ─────────────────────────────────────────────────────────────
+from crewai import Agent, Task, Crew, Process, LLM
+from crewai_tools import SerperDevTool, ScrapeWebsiteTool
+
+# ─── UI ──────────────────────────────────────────────────────────────────────
+import gradio as gr
+
+# =============================================================================
+# API KEYS — Read from Hugging Face Space Secrets
+# Go to: Space → Settings → Variables and secrets → New secret
+#   Secret name : OPENAI_API_KEY   → your OpenAI key
+#   Secret name : SERPER_API_KEY   → your Serper key (serper.dev)
+# =============================================================================
+OPENAI_API_KEY = os.environ.get("OPENAI_API_KEY", "")
+SERPER_API_KEY = os.environ.get("SERPER_API_KEY", "")
+
+if not OPENAI_API_KEY:
+    raise EnvironmentError(
+        "❌ OPENAI_API_KEY not found. "
+        "Add it in Space → Settings → Variables and secrets."
+    )
+
+if not SERPER_API_KEY:
+    raise EnvironmentError(
+        "❌ SERPER_API_KEY not found. "
+        "Add it in Space → Settings → Variables and secrets."
+    )
+
+# CrewAI and LangChain pick these up automatically from os.environ
+os.environ["OPENAI_API_KEY"] = OPENAI_API_KEY
+os.environ["SERPER_API_KEY"] = SERPER_API_KEY
+
+# =============================================================================
+# LLM
+# =============================================================================
+llm = LLM(
+    model="gpt-4o-mini",
+    temperature=0.4,
+    max_tokens=2000,
+)
+
+# =============================================================================
+# TOOLS
+# =============================================================================
+search_tool = SerperDevTool()       # Finds URLs + short snippets via Serper
+scrape_tool = ScrapeWebsiteTool()   # Reads full page content from those URLs
+
+# =============================================================================
+# MEMORY
+# =============================================================================
+MEMORY_DIR = "/tmp/crewai_memory"   # /tmp is always writable on HF Spaces
+os.makedirs(MEMORY_DIR, exist_ok=True)
+
+EMBEDDER_CONFIG = {
+    "provider": "openai",
+    "config": {
+        "model": "text-embedding-3-small",
+        "api_key": OPENAI_API_KEY,
+    },
+}
+
+# =============================================================================
+# AGENTS
+# =============================================================================
+researcher = Agent(
+    role="Senior Research Analyst",
+    goal=(
+        "Find the most accurate, up-to-date, and relevant information on the given topic. "
+        "Always prioritize credible sources, cite key statistics, and highlight emerging trends."
+    ),
+    backstory=(
+        "You are a veteran research analyst with 15 years of experience across technology, "
+        "business, and science journalism. You have a reputation for digging deep into topics, "
+        "cross-referencing multiple sources, and presenting findings in a clear, structured format. "
+        "You never make assumptions — everything you report is evidence-backed. "
+        "You organize research into clean bullet points with source context."
+    ),
+    tools=[search_tool, scrape_tool],   # Serper finds URLs, ScrapeWebsite reads full content
+    llm=llm,
+    verbose=True,
+    allow_delegation=False,
+    max_iter=5,
+)
+
+writer = Agent(
+    role="Professional Content Writer",
+    goal=(
+        "Transform the research summary into a compelling, well-structured, and audience-appropriate article. "
+        "Ensure the article flows naturally, uses clear headings, and maintains the requested tone throughout."
+    ),
+    backstory=(
+        "You are an award-winning content writer with expertise in long-form journalism, technical writing, "
+        "and digital publishing. You have written for Forbes, TechCrunch, and Harvard Business Review. "
+        "You have a talent for making complex topics accessible without dumbing them down. "
+        "You always start with a strong hook, use subheadings to guide the reader, "
+        "and end with a memorable conclusion that reinforces the key message."
+    ),
+    llm=llm,
+    verbose=True,
+    allow_delegation=False,
+    max_iter=4,
+)
+
+editor = Agent(
+    role="Senior Editor",
+    goal=(
+        "Review and refine the article to ensure it is grammatically flawless, tonally consistent, "
+        "logically structured, and ready for publication. Return only the final polished article."
+    ),
+    backstory=(
+        "You are a meticulous senior editor with 20 years of experience at top-tier publications. "
+        "You have an unmatched eye for inconsistency, passive voice overuse, redundant phrasing, "
+        "and weak transitions. You tighten prose without stripping its personality. "
+        "You ensure every paragraph earns its place and every sentence is purposeful. "
+        "You never add commentary — you deliver only the clean, final article."
+    ),
+    llm=llm,
+    verbose=True,
+    allow_delegation=False,
+    max_iter=3,
+)
+
+# =============================================================================
+# TASK BUILDER
+# =============================================================================
+def build_tasks(topic: str, audience: str, tone: str, length: str) -> list:
+    research_task = Task(
+        description=textwrap.dedent(f"""
+            Conduct thorough research on the following topic:
+            TOPIC: "{topic}"
+
+            Your research should be targeted for this audience: {audience}
+            Required depth: {length}
+
+            Your deliverables:
+            1. A 2-sentence topic overview
+            2. 5-8 key facts or statistics with source context
+            3. 3-5 recent developments or trends (2023-2025)
+            4. 2-3 expert opinions or notable perspectives
+            5. A list of 3 key takeaways
+
+            Format everything in clearly labelled bullet points.
+        """),
+        expected_output=(
+            "A comprehensive, well-structured research summary in bullet points "
+            "covering overview, key facts, recent developments, expert opinions, and takeaways."
+        ),
+        agent=researcher,
+    )
+
+    writing_task = Task(
+        description=textwrap.dedent(f"""
+            Using the research summary provided, write a complete article on:
+            TOPIC: "{topic}"
+
+            Specifications:
+            - Target Audience: {audience}
+            - Tone: {tone}
+            - Length: {length}
+
+            Structure:
+            1. Headline - compelling and specific
+            2. Introduction - hook + why this topic matters now
+            3. Section 1 with subheading - background / context
+            4. Section 2 with subheading - key developments or insights
+            5. Section 3 with subheading - implications or practical applications
+            6. Conclusion - summary + forward-looking statement
+
+            Rules:
+            - Do NOT use generic filler phrases like "In conclusion" or "It is worth noting"
+            - Use active voice whenever possible
+            - Every paragraph should add new value - no repetition
+        """),
+        expected_output=(
+            "A complete, structured article with headline, introduction, three body sections "
+            "with subheadings, and a conclusion - written in the requested tone for the specified audience."
+        ),
+        agent=writer,
+        context=[research_task],
+    )
+
+    editing_task = Task(
+        description=textwrap.dedent(f"""
+            You are the final gatekeeper before publication.
+            Review and edit the article for the following:
+
+            Checklist:
+            - Grammar and spelling are flawless
+            - Tone is consistent and matches: {tone}
+            - Audience appropriateness: {audience}
+            - No redundant sentences or paragraphs
+            - Transitions between sections are smooth
+            - Headline is punchy and specific
+            - Introduction hooks the reader immediately
+            - Conclusion ends with impact
+
+            IMPORTANT: Return ONLY the final polished article text.
+            Do NOT include your editorial notes, checklist, or any commentary.
+        """),
+        expected_output=(
+            "The final, publication-ready article - clean, polished, and complete. "
+            "No editor notes or commentary included."
+        ),
+        agent=editor,
+        context=[writing_task],
+    )
+
+    return [research_task, writing_task, editing_task]
+
+# =============================================================================
+# CREW RUNNER
+# =============================================================================
+def run_crew(topic: str, audience: str, tone: str, length: str) -> str:
+    if not topic or not topic.strip():
+        return "Please enter a topic before generating."
+
+    if len(topic.strip()) < 5:
+        return "Topic is too short. Please be more descriptive (e.g. 'AI in healthcare 2025')."
+
+    print(f"\n{'='*60}")
+    print(f"Starting CrewAI Pipeline")
+    print(f"   Topic    : {topic}")
+    print(f"   Audience : {audience}")
+    print(f"   Tone     : {tone}")
+    print(f"   Length   : {length}")
+    print(f"{'='*60}\n")
+
+    try:
+        tasks = build_tasks(topic, audience, tone, length)
+
+        crew = Crew(
+            agents=[researcher, writer, editor],
+            tasks=tasks,
+            process=Process.sequential,
+            verbose=True,
+            memory=True,
+            embedder=EMBEDDER_CONFIG,
+            memory_config={
+                "long_term": {
+                    "storage_path": MEMORY_DIR
+                }
+            },
+        )
+
+        print("\nAgent 1 (Researcher) starting...\n")
+        result = crew.kickoff()
+
+        print(f"\n{'='*60}")
+        print("CrewAI pipeline complete!")
+        print(f"{'='*60}\n")
+
+        return str(result)
+
+    except Exception as e:
+        error_msg = f"An error occurred during pipeline execution:\n{str(e)}"
+        print(error_msg)
+        return error_msg
+
+# =============================================================================
+# GRADIO UI
+# =============================================================================
+with gr.Blocks(
+    theme=gr.themes.Soft(primary_hue="blue", secondary_hue="slate"),
+    title="Multi-Agent Article Generator",
+    css="""
+        .title-text { text-align: center; }
+        .agent-badge { font-size: 0.85rem; color: #666; }
+        footer { display: none !important; }
+    """
+) as demo:
+
+    gr.Markdown(
+        """
+        <div class="title-text">
+        <h1>Multi-Agent Article Generator</h1>
+        <p style="color:#555;">Powered by <strong>CrewAI</strong> + <strong>GPT-4o</strong> + <strong>SerperDev</strong> + <strong>ScrapeWebsite</strong></p>
+        </div>
+
+        ---
+
+        ### How it works
+        `Researcher` searches the web + scrapes full pages for deep research →
+        `Writer` drafts a structured article →
+        `Editor` polishes and finalizes for publication
+
+        > Generation takes **45-120 seconds**. Agent logs appear in the Space logs panel.
+        """
+    )
+
+    with gr.Row(equal_height=False):
+
+        with gr.Column(scale=1, min_width=280):
+            gr.Markdown("### Settings")
+
+            topic_input = gr.Textbox(
+                label="Topic *",
+                placeholder="e.g. The impact of AI agents on software development in 2025",
+                lines=3,
+                info="Be specific - better topic = better article."
+            )
+
+            audience_input = gr.Dropdown(
+                label="Target Audience",
+                choices=[
+                    "General Public",
+                    "Business Professionals",
+                    "Students & Learners",
+                    "Researchers & Academics",
+                    "Tech Enthusiasts"
+                ],
+                value="General Public",
+                info="Shapes vocabulary and depth of explanation."
+            )
+
+            tone_input = gr.Dropdown(
+                label="Writing Tone",
+                choices=[
+                    "Informative",
+                    "Persuasive",
+                    "Conversational",
+                    "Academic",
+                    "Inspirational"
+                ],
+                value="Informative",
+                info="Controls the style and voice of the article."
+            )
+
+            length_input = gr.Radio(
+                label="Article Length",
+                choices=[
+                    "Short (300-500 words)",
+                    "Medium (600-900 words)",
+                    "Long (1000-1400 words)"
+                ],
+                value="Medium (600-900 words)",
+            )
+
+            gr.Markdown("---")
+
+            submit_btn = gr.Button(
+                "Generate Article",
+                variant="primary",
+                size="lg"
+            )
+
+            clear_btn = gr.Button(
+                "Clear",
+                variant="secondary",
+                size="sm"
+            )
+
+            gr.Markdown(
+                """
+                <div class="agent-badge">
+                <b>Agents:</b><br>
+                Researcher - web search + scrape full pages<br>
+                Writer - draft structured article<br>
+                Editor - polish + finalize
+                </div>
+                """
+            )
+
+        with gr.Column(scale=2):
+            gr.Markdown("### Generated Article")
+
+            output_box = gr.Textbox(
+                label="",
+                lines=35,
+                placeholder=(
+                    "Your article will appear here once generation is complete.\n\n"
+                    "This takes 45-120 seconds.\n"
+                    "Watch the Space logs for live agent output."
+                ),
+                show_copy_button=True,
+                interactive=False,
+            )
+
+    gr.Markdown("### Example Topics")
+    gr.Examples(
+        examples=[
+            ["The future of AI agents in enterprise software", "Business Professionals", "Informative", "Medium (600-900 words)"],
+            ["How large language models are changing education", "Students & Learners", "Conversational", "Short (300-500 words)"],
+            ["Quantum computing: where are we in 2025?", "Tech Enthusiasts", "Informative", "Long (1000-1400 words)"],
+            ["The ethics of autonomous AI decision-making", "Researchers & Academics", "Academic", "Long (1000-1400 words)"],
+            ["Why every startup needs an AI strategy in 2025", "Business Professionals", "Persuasive", "Medium (600-900 words)"],
+        ],
+        inputs=[topic_input, audience_input, tone_input, length_input],
+        label="Click any example to auto-fill the inputs"
+    )
+
+    gr.Markdown(
+        """
+        ---
+        <div style="text-align:center; color:#888; font-size:0.85rem;">
+        Built with <strong>CrewAI</strong> · <strong>LangChain</strong> · <strong>GPT-4o</strong> · <strong>Gradio</strong> · <strong>ScrapeWebsiteTool</strong>
+        </div>
+        """
+    )
+
+    submit_btn.click(
+        fn=run_crew,
+        inputs=[topic_input, audience_input, tone_input, length_input],
+        outputs=output_box,
+        api_name="generate",
+    )
+
+    clear_btn.click(
+        fn=lambda: ("", ""),
+        inputs=[],
+        outputs=[topic_input, output_box]
+    )
+
+# =============================================================================
+# LAUNCH
+# server_name="0.0.0.0" is required for Hugging Face Spaces
+# server_port=7860 is the default port HF Spaces expects
+# =============================================================================
+demo.launch(server_name="0.0.0.0", server_port=7860)