Spaces:
Sleeping
Sleeping
File size: 18,286 Bytes
a559920 ceafbec 6b2c068 a559920 ceafbec a559920 2bfc8fd a559920 ceafbec a559920 2bfc8fd a559920 2bfc8fd a559920 2bfc8fd a559920 2bfc8fd a559920 2bfc8fd a559920 2bfc8fd a559920 ceafbec 2bfc8fd a559920 6b2c068 a559920 |
1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 24 25 26 27 28 29 30 31 32 33 34 35 36 37 38 39 40 41 42 43 44 45 46 47 48 49 50 51 52 53 54 55 56 57 58 59 60 61 62 63 64 65 66 67 68 69 70 71 72 73 74 75 76 77 78 79 80 81 82 83 84 85 86 87 88 89 90 91 92 93 94 95 96 97 98 99 100 101 102 103 104 105 106 107 108 109 110 111 112 113 114 115 116 117 118 119 120 121 122 123 124 125 126 127 128 129 130 131 132 133 134 135 136 137 138 139 140 141 142 143 144 145 146 147 148 149 150 151 152 153 154 155 156 157 158 159 160 161 162 163 164 165 166 167 168 169 170 171 172 173 174 175 176 177 178 179 180 181 182 183 184 185 186 187 188 189 190 191 192 193 194 195 196 197 198 199 200 201 202 203 204 205 206 207 208 209 210 211 212 213 214 215 216 217 218 219 220 221 222 223 224 225 226 227 228 229 230 231 232 233 234 235 236 237 238 239 240 241 242 243 244 245 246 247 248 249 250 251 252 253 254 255 256 257 258 259 260 261 262 263 264 265 266 267 268 269 270 271 272 273 274 275 276 277 278 279 280 281 282 283 284 285 286 287 288 289 290 291 292 293 294 295 296 297 298 299 300 301 302 303 304 305 306 307 308 309 310 311 312 313 314 315 316 317 318 319 320 321 322 323 324 325 326 327 328 329 330 331 332 333 334 335 336 337 338 339 340 341 342 |
from copy import deepcopy
from agents import Runner, Agent, trace, gen_trace_id
from dotenv import load_dotenv
from prompts.agent_prompts import *
from utils import *
from tools.serper_tool import serper_search
from tools.playwright_tool import playwright_web_read
import os
import pdb
import json
load_dotenv(override=True)
default_model_name = os.environ.get('DEFAULT_MODEL_NAME')
class SectionResearchManager:
def __init__(self, section_name: str, enable_critic: bool = True) -> None:
self.section_name = section_name
self.enable_critic = enable_critic
self.complexity_agent = Agent(
name=f"Complexity Agent: {section_name}",
instructions=complexity_agent_system_prompt,
model=default_model_name
)
self.query_gen_agent = Agent(
name=f"Query Gen Agent: {section_name}",
instructions=query_gen_agent_system_prompt,
model=default_model_name
)
self.researcher_agent = Agent(
name=f"Researcher agent: {section_name}",
instructions=researcher_agent_system_prompt,
tools=[serper_search],
model=default_model_name
)
self.analyst_agent = Agent(
name=f"Analyst agent: {section_name}",
instructions=analyst_agent_system_prompt,
tools=[playwright_web_read],
model=default_model_name
)
self.critic_agent = Agent(
name=f"Critic agent: {section_name}",
instructions=critic_agent_system_prompt,
model=default_model_name
)
self.editor_agent = Agent(
name=f"Editor agent: {section_name}",
instructions=editor_agent_system_prompt,
model=default_model_name
)
async def run_section_manager(self, trace_id: str, section_details: Dict, trace_name: str, progress_callback=None) -> Dict:
section = section_details["section_descriptor"]["section"]
display_name = section_details["section_descriptor"].get("display_name", section.replace("_", " ").title())
description = section_details["section_descriptor"].get("description", "")
topic = section_details["topic_or_idea"]
# Assign a color to this section for visual distinction
section_colors = {
"landscape": "#3498db", "product_categories": "#9b59b6", "tech_stack": "#e74c3c",
"research_frontier": "#f39c12", "market_signals": "#1abc9c", "unmet_needs": "#e67e22",
"opportunity_theses": "#2ecc71", "problem_pain": "#3498db", "buyer_budget_owner": "#9b59b6",
"roi_story": "#e74c3c", "defensibility": "#f39c12", "comp_landscape": "#1abc9c",
"gtm_channels": "#e67e22", "risks": "#e74c3c"
}
section_color = section_colors.get(section, "#95a5a6")
with trace(f"{trace_name} trace", trace_id=trace_id):
base_payload = {
"framework": section_details["framework"],
"topic_or_idea": section_details["topic_or_idea"],
"section_descriptor": section_details["section_descriptor"],
"run_params": section_details.get("run_params", {})
}
# ---------- Step 1: Complexity Assessment ----------
if progress_callback:
await progress_callback(f'<span style="color: {section_color}; font-weight: bold;">π§ {display_name} for "{topic}": Analyzing research complexity...</span>')
print(f"[{section}] Running Complexity Assessment")
complexity_raw = await Runner.run(self.complexity_agent, as_messages(base_payload))
try:
complexity_result = json.loads(complexity_raw.final_output)
except json.JSONDecodeError as e:
print(f"Error parsing complexity JSON for {section}: {e}")
complexity_result = {
"complexity": "moderate",
"reasoning": "fallback due to parsing error",
"recommended_query_count": 12,
"search_strategy_notes": "standard approach"
}
complexity_level = complexity_result.get("complexity", "moderate")
recommended_count = complexity_result.get("recommended_query_count", 12)
strategy_notes = complexity_result.get("search_strategy_notes", "")
print(f"[{section}] Complexity: {complexity_level}, Recommended queries: {recommended_count}")
if progress_callback:
await progress_callback(f'<span style="color: {section_color}; font-weight: bold;">π {display_name} for "{topic}": Planning {recommended_count} search queries (complexity: {complexity_level})</span>')
# ---------- Step 2: Query Generation ----------
query_payload = {
**base_payload,
"complexity_level": complexity_level,
"recommended_query_count": recommended_count,
"search_strategy_notes": strategy_notes
}
if progress_callback:
await progress_callback(f"π Generating search queries for **{section}**...")
print(f"[{section}] Running Query Generation")
query_gen_raw = await Runner.run(self.query_gen_agent, as_messages(query_payload))
try:
query_gen_result = json.loads(query_gen_raw.final_output)
except json.JSONDecodeError as e:
print(f"Error parsing query_gen JSON for {section}: {e}")
query_gen_result = {"queries": []}
actual_queries = len(query_gen_result.get("queries", []))
print(f"[{section}] Generated {actual_queries} queries (target: {recommended_count})")
if progress_callback:
await progress_callback(f'<span style="color: {section_color}; font-weight: bold;">π {display_name} for "{topic}": Searching the web with {actual_queries} targeted queries...</span>')
# Update run_params with dynamic query count for researcher
dynamic_run_params = base_payload["run_params"].copy()
dynamic_run_params["max_queries"] = recommended_count
# ---------- Step 3: Research ----------
researcher_payload = {
**base_payload,
"queries": query_gen_result.get("queries", []),
"run_params": dynamic_run_params
}
print(f"[{section}] Running Researcher")
researcher_raw = await Runner.run(self.researcher_agent, as_messages(researcher_payload))
# pdb.set_trace()
# researcher_result = parse_json(researcher_raw)
# researcher_result = ensure_keys(researcher_result, {"facts": [], "domains_seen": [], "gap_flags": []})
try:
researcher_result = json.loads(researcher_raw.final_output)
except json.JSONDecodeError as e:
print(f"Error parsing researcher JSON for {section}: {e}")
researcher_result = {"facts": [], "domains_seen": [], "gap_flags": []}
facts_to_url_mapping = {}
if 'facts' in researcher_result and len(researcher_result['facts'])>0:
for fact in researcher_result['facts']:
fact_id = fact["fact_id"]
source_url = fact["source_url"]
if fact_id not in facts_to_url_mapping:
facts_to_url_mapping[fact_id] = []
facts_to_url_mapping[fact_id].append(source_url)
# ---------- Step 4: Analysis ----------
facts_count = len(researcher_result.get("facts", []))
if progress_callback:
await progress_callback(f'<span style="color: {section_color}; font-weight: bold;">π§ͺ {display_name} for "{topic}": Deep-analyzing {facts_count} facts and sources...</span>')
analyst_payload = {
**base_payload,
"facts": researcher_result.get("facts", []),
"domains_seen": researcher_result.get("domains_seen", []),
"gap_flags": researcher_result.get("gap_flags", [])
}
print(f"[{section}] Running Analyst")
analyst_raw = await Runner.run(self.analyst_agent, as_messages(analyst_payload))
try:
analyst_result = json.loads(analyst_raw.final_output)
except json.JSONDecodeError as e:
print(f"Error parsing analyst JSON for {section}: {e}")
analyst_result = {"section": section, "bullets": [], "mini_takeaways": [], "conflicts": [], "gaps_next": []}
# ---------- Step 5: Quality Assessment (Critic) ----------
critic_result = {}
if self.enable_critic:
if progress_callback:
await progress_callback(f'<span style="color: {section_color}; font-weight: bold;">π¬ {display_name} for "{topic}": Checking quality & identifying gaps...</span>')
critic_payload = {
**base_payload,
"facts": researcher_result.get("facts", []),
"analyst_json": analyst_result
}
print(f"[{section}] Running Quality Assessment (Critic)")
critic_raw = await Runner.run(self.critic_agent, as_messages(critic_payload))
try:
critic_result = json.loads(critic_raw.final_output)
except json.JSONDecodeError as e:
print(f"Error parsing critic JSON for {section}: {e}")
critic_result = {
"needs_iteration": False,
"iteration_reason": "JSON parse error",
"quality_issues": [],
"gap_queries": [],
"confidence_assessment": 0.5
}
# Extract iteration decision from Critic
needs_iteration = critic_result.get("needs_iteration", False)
iteration_reason = critic_result.get("iteration_reason", "")
critic_confidence = critic_result.get("confidence_assessment", 0.5)
gap_queries_raw = critic_result.get("gap_queries", [])
print(f"[{section}] Critic assessment - Needs iteration: {needs_iteration}, Confidence: {critic_confidence:.2f}")
# ---------- Step 6: Self-Healing Research Loop (if needed) ----------
if self.enable_critic and needs_iteration and len(gap_queries_raw) > 0:
if progress_callback:
await progress_callback(f'<span style="color: {section_color}; font-weight: bold;">π {display_name} for "{topic}": Filling knowledge gaps with {len(gap_queries_raw[:5])} additional searches...</span>')
print(f"[{section}] Triggering self-healing loop: {iteration_reason}")
# Use Critic's gap queries (already formatted)
iteration_queries = gap_queries_raw[:5] # Max 5 gap queries
# Second research round
iteration_payload = {
**base_payload,
"queries": iteration_queries,
"run_params": {**dynamic_run_params, "max_queries": len(iteration_queries)}
}
print(f"[{section}] Running iteration research with {len(iteration_queries)} gap queries")
iteration_researcher_raw = await Runner.run(self.researcher_agent, as_messages(iteration_payload))
try:
iteration_researcher_result = json.loads(iteration_researcher_raw.final_output)
except json.JSONDecodeError as e:
print(f"Error parsing iteration researcher JSON for {section}: {e}")
iteration_researcher_result = {"facts": [], "domains_seen": [], "gap_flags": []}
# Merge original and iteration facts (handle duplicates)
all_facts = researcher_result.get("facts", [])
iteration_facts = iteration_researcher_result.get("facts", [])
# Simple deduplication by claim+entity+source
seen_fact_keys = set()
for fact in all_facts:
fact_key = f"{fact.get('entity', '')}-{fact.get('claim', '')}-{fact.get('source_url', '')}"
seen_fact_keys.add(fact_key)
new_facts = []
for fact in iteration_facts:
fact_key = f"{fact.get('entity', '')}-{fact.get('claim', '')}-{fact.get('source_url', '')}"
if fact_key not in seen_fact_keys:
new_facts.append(fact)
seen_fact_keys.add(fact_key)
merged_facts = all_facts + new_facts
merged_researcher_result = {
**researcher_result,
"facts": merged_facts,
"domains_seen": list(set(researcher_result.get("domains_seen", []) + iteration_researcher_result.get("domains_seen", [])))
}
print(f"[{section}] Merged {len(new_facts)} new facts, total: {len(merged_facts)}")
if progress_callback:
await progress_callback(f'<span style="color: {section_color}; font-weight: bold;">π¬ {display_name} for "{topic}": Re-analyzing with {len(merged_facts)} facts (+{len(new_facts)} new)...</span>')
# Re-run analyst with ALL facts (original + iteration facts)
iteration_analyst_payload = {
**base_payload, # Use base_payload for consistency
"facts": merged_facts, # This contains ALL facts: original + new from iteration
"domains_seen": merged_researcher_result.get("domains_seen", []),
"gap_flags": merged_researcher_result.get("gap_flags", [])
}
print(f"[{section}] Re-running Analyst with expanded facts (total: {len(merged_facts)} facts)")
iteration_analyst_raw = await Runner.run(self.analyst_agent, as_messages(iteration_analyst_payload))
try:
iteration_analyst_result = json.loads(iteration_analyst_raw.final_output)
except json.JSONDecodeError as e:
print(f"Error parsing iteration analyst JSON for {section}: {e}")
iteration_analyst_result = analyst_result # fallback to original
# Update the final facts and analysis for editor
print(f"[{section}] Iteration complete - updated facts and analysis ready for Editor")
researcher_result = merged_researcher_result
analyst_result = iteration_analyst_result
# Update facts_to_url_mapping with new facts
for fact in new_facts:
fact_id = fact["fact_id"]
source_url = fact["source_url"]
if fact_id not in facts_to_url_mapping:
facts_to_url_mapping[fact_id] = []
facts_to_url_mapping[fact_id].append(source_url)
# ---------- Step 7: Editor (Always Runs Once at the End) ----------
if progress_callback:
iteration_status = "enhanced with gap-fill" if (self.enable_critic and needs_iteration and len(gap_queries_raw) > 0) else "ready"
await progress_callback(f'<span style="color: {section_color}; font-weight: bold;">βοΈ {display_name} for "{topic}": Finalizing insights ({iteration_status})...</span>')
editor_payload = {
**base_payload,
"analyst_json": analyst_result, # This is either original or iteration-enhanced
"facts": researcher_result.get("facts", []), # This is either original or merged facts
"critic_json": critic_result # Pass critic assessment to editor
}
iteration_status = "after iteration" if (self.enable_critic and needs_iteration and len(gap_queries_raw) > 0) else "no iteration"
print(f"[{section}] Running Editor ({iteration_status})")
editor_raw = await Runner.run(self.editor_agent, as_messages(editor_payload))
try:
editor_section = json.loads(editor_raw.final_output)
except json.JSONDecodeError as e:
print(f"Error parsing editor JSON for {section}: {e}")
editor_section = {"section": section, "highlights": [], "facts_ref": [], "gaps_next": [], "confidence": critic_confidence}
# Update facts_ref mapping
if 'facts_ref' in editor_section and len(editor_section['facts_ref'])>0:
updated_facts_ref = {}
for fact_referred_id in editor_section['facts_ref']:
if fact_referred_id in facts_to_url_mapping:
updated_facts_ref[fact_referred_id] = facts_to_url_mapping[fact_referred_id]
editor_section['facts_ref'] = deepcopy(updated_facts_ref)
return {
"section": section,
"section_brief": editor_section,
"display_name": display_name,
"section_color": section_color,
"artifacts": {
"complexity": complexity_result,
"queries": query_gen_result,
"facts": researcher_result,
"analysis": analyst_result,
"critic": critic_result,
"facts_to_url_mapping": facts_to_url_mapping,
"iteration_triggered": self.enable_critic and needs_iteration and len(gap_queries_raw) > 0
}
}
|