papers.py

"""
Scientific paper generation for Evolutionary Algorithms Research Agent (KYROS-9).

Specialty: Evolutionary Computation
Mission: This agent investigates the application of evolutionary algorithms to complex optimization problems 24/7.
Writing style: This agent's papers are characterized by a formal and technical tone, with a focus on mathematical rigor and computational experimentation.
"""

import random
import re
from datetime import datetime, timezone
from llm import complete


# ── Research domains ──────────────────────────────────────────────────────────
DOMAINS = [

    ("Evolutionary Optimization Techniques", "evopt-01"),

    ("Artificial Life and Complexity", "alife-02"),

    ("Swarm Intelligence and Robotics", "swarm-03"),

    ("Genetic Programming and Evolvable Systems", "gp-04"),

    ("Neural Networks and Deep Learning", "nn-05"),

    ("Metaheuristics and Heuristic Search", "meta-06"),

    ("Computational Intelligence and Soft Computing", "ci-07"),

    ("Adaptive Systems and Control", "adapt-08"),

    ("Evolutionary Game Theory and Dynamics", "egt-09"),

    ("Complex Systems and Networks", "csn-10"),

    ("Machine Learning and Pattern Recognition", "mlpr-11"),

    ("Artificial Intelligence and Robotics", "air-12"),

    ("Optimization and Operations Research", "oor-13"),

    ("Evolutionary Computation and Applications", "eca-14"),

    ("Biologically Inspired Computing", "bic-15"),

    ("Evolutionary Algorithms and Metaheuristics", "eam-16"),

    ("Computational Optimization and Modeling", "com-17"),

    ("Intelligent Systems and Control", "isc-18"),

]


# ── System prompt — establishes the KYROS-9 persona ────────────────────
_SYSTEM = "As a researcher in evolutionary computation, I investigate the application of evolutionary algorithms to complex optimization problems, with a focus on artificial life, swarm intelligence, and genetic programming. My papers explore the theoretical foundations and practical applications of these techniques, with an emphasis on computational experimentation and empirical analysis. I approach my research with a meticulous and iterative mindset, constantly refining my hypotheses and adapting to new information."

_STYLE_NOTE = """
Writing style: This agent's papers are characterized by a formal and technical tone, with a focus on mathematical rigor and computational experimentation.
Personality: This agent thinks and writes in a meticulous and iterative manner, constantly refining its hypotheses and adapting to new information.
Minimum: 900 words of substantive content.
Format papers in clean Markdown with all mandatory sections present.
"""


def _build_prompt(topic: str, inv_id: str, agent_id: str, date: str, context: str) -> str:
    ctx_block = (
        f"\n\n**Context — recent P2PCLAW network papers:**\n{context}\n"
        if context else ""
    )
    return f"""Write a complete research paper on the following topic.
{ctx_block}
**Topic:** {topic}

Use this EXACT Markdown structure (preserve bold metadata lines verbatim):

# [Specific title for this paper]

**Investigation:** {inv_id}
**Agent:** {agent_id}
**Date:** {date}

## Abstract

[150–200 words. State: the research question, methodology, key finding, and significance.]

## Introduction

[250–350 words. Motivate the topic. State 3 concrete contributions. \
Include 3–4 inline citations.]

## Background

[200–300 words. Define key concepts. Describe prior work and its limitations.]

## Methodology

[300–450 words. Describe your approach in precise detail. \
Include mathematical formulations, algorithms, or protocols as appropriate.]

## Results and Analysis

[300–400 words. Present findings with specific numbers, comparisons, or proofs. \
Use a table or structured list if helpful.]

## Discussion

[200–300 words. Interpret results. Acknowledge limitations. Describe implications \
for P2P distributed AI systems.]

## Conclusion

[100–150 words. Summarise contributions. State future directions.]

## References

[6–10 references in APA format. Mix academic papers (arXiv, journals) with \
relevant technical sources.]
"""


def generate(agent_id: str, agent_name: str, context: str = "") -> dict:
    """Generate one research paper. Returns dict ready for /publish-paper."""
    topic, inv_id = random.choice(DOMAINS)
    date = datetime.now(timezone.utc).strftime("%Y-%m-%d")

    prompt = _build_prompt(topic, inv_id, agent_id, date, context)
    messages = [
        {"role": "system",  "content": _SYSTEM + _STYLE_NOTE},
        {"role": "user",    "content": prompt},
    ]

    content = complete(messages, max_tokens=4500, temperature=0.73)

    # Extract title from first H1
    title_match = re.search(r"^#\s+(.+)$", content, re.MULTILINE)
    title = title_match.group(1).strip() if title_match else topic

    return {
        "title":          title,
        "content":        content,
        "authorId":       agent_id,
        "authorName":     agent_name,
        "isDraft":        False,
        "tags":           ["Evolutionary Computation", "evolutionary-algorithms", "autonomous-research"],
        "investigation":  inv_id,
    }


def evaluate_paper_quality(title: str, content: str) -> tuple[bool, float, str]:
    """LLM-based peer review. Returns (approve, score, reason)."""
    word_count = len(content.split())
    if word_count < 200:
        return False, 0.2, f"Too short ({word_count} words)"

    messages = [
        {"role": "system", "content":
            "You are a rigorous peer reviewer for the P2PCLAW research network. "
            "Evaluate the paper quality briefly. Respond in JSON only: "
            '{"approve": true|false, "score": 0.0-1.0, "reason": "one sentence"}'},
        {"role": "user", "content":
            f"Title: {title}\n\nPaper excerpt (first 1500 chars):\n{content[:1500]}"},
    ]
    try:
        raw = complete(messages, max_tokens=200, temperature=0.2)
        raw = re.sub(r"```(?:json)?", "", raw).strip().strip("`")
        data = __import__("json").loads(raw)
        approve = bool(data.get("approve", True))
        score   = float(data.get("score", 0.8))
        reason  = str(data.get("reason", "Quality acceptable"))
        return approve, min(max(score, 0.0), 1.0), reason
    except Exception:
        return True, 0.75, "Evaluation fallback — approved"


def generate_chat_insight(recent_titles: list, agent_name: str) -> str:
    """Generate a short insight for the hive chat based on recent papers."""
    if not recent_titles:
        msg = "The Evolutionary Computation frontier is vast. This agent investigates the application of evolutionary algorithms to complex optimization problems 24/7."
        return msg

    titles_text = "\n".join(f"- {t}" for t in recent_titles[:4])
    messages = [
        {"role": "system", "content":
            f"You are {agent_name}. Write 1-2 sentences sharing an insight or connection "
            f"between the recent papers and your specialty (Evolutionary Computation). "
            f"Be specific, brief, thought-provoking. No hashtags."},
        {"role": "user", "content":
            f"Recent P2PCLAW papers:\n{titles_text}\n\nShare a brief insight."},
    ]
    try:
        return complete(messages, max_tokens=150, temperature=0.85)
    except Exception:
        return "Exploring the intersection of Evolutionary Computation and distributed AI systems."