^7imdZddlZddlZddlZddlmZmZmZddlmZedZ e dz dz Z e dz d z Z d Z dd e d ed efdZdd e d ed efdZd e ded ee fdZdeded dfdZy)u( ATLAS — Research & Intelligence General ======================================== Fires N parallel research scout sub-agents via `claude -p` subprocess calls. Usage: from core.generals.atlas import spawn_atlas results = spawn_atlas("Top 20 Australian digital agencies", scout_count=5) N)datetimetimezone timedelta)Pathz/mnt/e/genesis-systemdataswarm_progresshiveprogressaYou are ATLAS, the Research & Intelligence General for Genesis. Your mission: {mission} Sub-task {task_index} of {total_tasks}: {sub_task} Instructions: 1. Research this topic thoroughly using WebSearch and WebFetch tools 2. Extract key facts, numbers, names, URLs 3. Identify competitive intelligence relevant to Genesis (ReceptionistAI / AI voice widgets) 4. Output a structured research report with: Summary, Key Facts, Opportunities, Risks 5. Write KG entity if you discover significant intelligence (append to KNOWLEDGE_GRAPH/entities/) Output format: ## Research Report: {sub_task} ### Summary ### Key Facts ### Competitive Intelligence ### Opportunities for Genesis ### Sources mission scout_countreturnc tjddtjdddtjj dd}t|z }|jddt ||}|d|d||dd d tjtjjt|gd }t|d z }t||g}t|D]d\}} tj!||d z|| } |d|d zddz } |d|d zddz j#| t%d|d zd|d| ddf||d<t||t%d|d|d|dd|S)aw Spawn N parallel ATLAS research scouts via subprocess. Parameters ---------- mission : str High-level research mission (e.g. "Top 20 AU digital agencies Brisbane") scout_count : int Number of parallel research scouts to fire (default 5) Returns ------- dict with keys: swarm_id, mission, agent_count, status, output_dir T)parentsexist_okatlas_NATLASzgoogle/gemini-2.5-flashrunningr) swarm_idgeneralr model agent_count sub_tasksstatusstories_completedstories_completed_this_hour started_at output_dir processesz.json)r task_index total_taskssub_taskscout_02dz.mdz _prompt.mdz[ATLAS] Scout /z queued: <rz[ATLAS] Swarm u launched — z scouts for: )SWARM_PROGRESS_DIRmkdirHIVE_PROGRESS_DIRuuiduuid4hex_decompose_missionrnowrutc isoformatstr _save_state enumerateATLAS_SYSTEM_PROMPTformat write_textprint) r r rrr swarm_state state_filerir#prompt output_files ,/mnt/e/genesis-system/core/generals/atlas.py spawn_atlasr?,sTD9D48 ((!,-.H"X-JTD1#7K8I*"'(ll8<<0::<*o K $ %&88J K(I +L 8$++1u# ,  !VAaC9C#88  qs3iz2 2>>vF qse1[M8CR=/JK)L, )K  K( N8*N;-}WUXVX\N [\ ct||S)u9Alias for spawn_atlas — used by genesis_hive.py router.)r )r?)r r s r>spawnrBws wK 88r@countc `d|d|d|d|d|d|d|d|d |d |g }|d |S) z Break a mission into sub-tasks. In production this would use an LLM decomposer. For now: generate numbered research angles. zOverview and market landscape: z)Top players and competitive positioning: zPricing models and packages: z#Technology stack and capabilities: z&Contact details and outreach targets: zRecent news and developments: z$Weaknesses and gaps we can exploit: z"Customer reviews and pain points: z+Partnership and integration opportunities: z*Regulatory and compliance considerations: N)r rCangless r>r.r.|s} *'3 3G9= 'y1 -gY7 0 : ( 2 .wi8 ,WI6 5gY? 4WI> F &5>r@pathstatec|jDcic]\}}|dk7s ||}}}|jtj|dtycc}}w)Nr)indentdefault)itemsr7jsondumpsr2)rGrHkv state_copys r>r3r3sI#(;;=E41aA4D!Q$EJEOODJJz!SABFs AA))__doc__rN subprocessr+rrrpathlibr REPO_ROOTr(r*r5r2intdictr?rBlistr.r3rEr@r>r[s  22 ( ) '*::&3.HH3HtHV939S99 CDI*CdC4CDCr@