iVbdZddlZddlZddlZddlZddlZddlmZmZm Z m Z ddl m Z mZddlmZddlmZejeZgdZdZd Zd Zd Zd Zed dZdede edefdZeddfdedede de!de!de"eee#ff dZ$edfdedede de!def dZ% djdede e ede edefdZ&de#fd Z'deeeffd!Z(ed"kr e)d#e)d$eZ*e*s-e)d%e)d&e)d'e+d n e)d(e)d)e(Z,e)d*e)d+e-e,dd,d-.e,de)d.e,de)d/e,d0e)d1e,d2e)d3e,d4e)d5e,d6ej/r(e)d7ej0e)d8ej0d9n e)d:e)d;e)d<e)d=e)d>e)d?e)d@e)dAe)dBe)dCe)dDe)dEe)dFe)dGe)dHe)dIe)dJe)dKe)dLe)dMe)dNe)dOedPedQe)dRe)dSe)dTe)dUe)dVe)dWe)dXe)dYe)dZdd[l1m2Z2d\d]d^dd_d`daidgdbdcZ3e2j4d\dde3dee'dfgdgdhidS)kaQ Mixture-of-Agents Tool Module This module implements the Mixture-of-Agents (MoA) methodology that leverages the collective strengths of multiple LLMs through a layered architecture to achieve state-of-the-art performance on complex reasoning tasks. Based on the research paper: "Mixture-of-Agents Enhances Large Language Model Capabilities" by Junlin Wang et al. (arXiv:2406.04692v1) Key Features: - Multi-layer LLM collaboration for enhanced reasoning - Parallel processing of reference models for efficiency - Intelligent aggregation and synthesis of diverse responses - Specialized for extremely difficult problems requiring intense reasoning - Optimized for coding, mathematics, and complex analytical tasks Available Tool: - mixture_of_agents_tool: Process complex queries using multiple frontier models Architecture: 1. Reference models generate diverse initial responses in parallel 2. Aggregator model synthesizes responses into a high-quality output 3. Multiple layers can be used for iterative refinement (future enhancement) Models Used (via OpenRouter): - Reference Models: claude-opus-4.6, gemini-3-pro-preview, gpt-5.4-pro, deepseek-v3.2 - Aggregator Model: claude-opus-4.6 (highest capability for synthesis) Configuration: To customize the MoA setup, modify the configuration constants at the top of this file: - REFERENCE_MODELS: List of models for generating diverse initial responses - AGGREGATOR_MODEL: Model used to synthesize the final response - REFERENCE_TEMPERATURE/AGGREGATOR_TEMPERATURE: Sampling temperatures - MIN_SUCCESSFUL_REFERENCES: Minimum successful models needed to proceed Usage: from mixture_of_agents_tool import mixture_of_agents_tool import asyncio # Process a complex query result = await mixture_of_agents_tool( user_prompt="Solve this complex mathematical proof..." ) N)DictAnyListOptional)get_async_client check_api_key)extract_content_or_reasoning) DebugSession)anthropic/claude-opus-4.6zgoogle/gemini-2.5-prozopenai/gpt-5.4-prozdeepseek/deepseek-v3.2r g333333?g?a^You have been provided with a set of responses from various open-source models to the latest user query. Your task is to synthesize these responses into a single, high-quality response. It is crucial to critically evaluate the information provided in these responses, recognizing that some of it may be biased or incorrect. Your response should not simply replicate the given answers but should offer a refined, accurate, and comprehensive reply to the instruction. Ensure your response is well-structured, coherent, and adheres to the highest standards of accuracy and reliability. Responses from models: moa_toolsMOA_TOOLS_DEBUG)env_var system_prompt responsesreturnchddt|D}|d|S)aG Construct the final system prompt for the aggregator including all model responses. Args: system_prompt (str): Base system prompt for aggregation responses (List[str]): List of responses from reference models Returns: str: Complete system prompt with enumerated responses  c(g|]\}}|dzd|S)r z. ).0iresponses A/home/ubuntu/.hermes/hermes-agent/tools/mixture_of_agents_tool.py z0_construct_aggregator_prompt..ds.\\\ 8!A#3333\\\z )join enumerate)rr response_texts r_construct_aggregator_promptr YsAII\\yQZG[G[\\\]]M 0 0 0 00ri}model user_prompt temperature max_tokens max_retriescKt|D]t} td||dz||d|dg|ddddid }|d s||d <t jjjdi|d {V}t|}|sXt d ||dz|||dz kr/tj td|dzzdd {Vtd|t|||dfcS#t$rF} t!| } d| vr!t d||dz| nmd| vsd| vr!t d||dz| n t d||dz| ||dz krLtd|dzzd} td| tj | d {Vn3|d|d| } td| d|| dfcYd } ~ cSYd } ~ nd } ~ wwxYwd S)a Run a single reference model with retry logic and graceful failure handling. Args: model (str): Model identifier to use user_prompt (str): The user's query temperature (float): Sampling temperature for response generation max_tokens (int): Maximum tokens in response max_retries (int): Maximum number of retry attempts Returns: tuple[str, str, bool]: (model_name, response_content_or_error, success_flag) zQuerying %s (attempt %s/%s)r userrolecontent reasoningTxhighenabledeffortr"messagesr% extra_bodygpt-r$Nz3%s returned empty content (attempt %s/%s), retrying<z%s responded (%s characters)invalidz)%s invalid request error (attempt %s): %sratelimitz$%s rate limit error (attempt %s): %sz!%s unknown error (attempt %s): %szRetrying in %ss...z failed after z attempts: %sexc_infoFr)rangeloggerinfolower startswith_get_openrouter_clientchat completionscreater warningasynciosleepminlen Exceptionstrerror) r"r#r$r%r&attempt api_paramsrr+e error_str sleep_time error_msgs r_run_reference_model_saferThs,(%%5/5/4 / KK5ugk; W W W&,EEF(#'")""  J;;==++F33 8,7 =)M355:FM[[PZ[[[[[[[[H28<(C(CDDDDDDDDD KK6s7|| L L L'4' ' ' ' / / /AIIOO----JESZ]^S^`ijjjj9??,,,,9??;L;L0L0LEugXYk[deeeeBE7UV;Xabbbq(( w{!3R88  0*===mJ//////////$WWKWWIWW  T9t <<<i......... 0//// /E5/5/s%C-D4-D44 J?D3J2JJcFKtdttd|dd|dg|ddddid }td s||d <t jjjdi|d {V}t|}|sRt d t jjjdi|d {V}t|}tdt||S)a Run the aggregator model to synthesize the final response. Args: system_prompt (str): System prompt with all reference responses user_prompt (str): Original user query temperature (float): Focused temperature for consistent aggregation max_tokens (int): Maximum tokens in final response Returns: str: Synthesized final response zRunning aggregator model: %ssystemr)r(r,Tr-r.r1r4r$Nz0Aggregator returned empty content, retrying oncez$Aggregation complete (%s characters)r) r>r?AGGREGATOR_MODELr@rArBrCrDrEr rFrJ)rr#r$r%rOrr+s r_run_aggregator_modelrXsZ$ KK.0@AAA"- 8 8 4 4 ! !   J"  ! ! # # . .v 6 60$/ =!E+--2>ESS SSSSSSSSH*844G 9IJJJI/116BIWWJWWWWWWWW.x88 KK6G EEE Nrreference_modelsaggregator_modelc \Ktj}tdkr dddzn|pt|ptt t tdddddgddid } t dt d dd tj d std |pt}|pt}t d t|t dtjfd|Dd{V}g}g} |D]3\} } } | r|| | | 4t|} t| }t d| || r.tdd| | tkr+td| dt|dtd| |d<||d<| |d<t dt%t&|}t)|t d{V}tj}||z }t d|d|||dd}d|d<t||d <||d!<|d"|d"<t,d#|t,t3j|d$d%S#t6$r}d&t9|}td'|d(tj}||z }dd)|pt|ptd|d*}||d+<||d!<t,d#|t,t3j|d$d%cYd}~Sd}~wwxYw),a Process a complex query using the Mixture-of-Agents methodology. This tool leverages multiple frontier language models to collaboratively solve extremely difficult problems requiring intense reasoning. It's particularly effective for: - Complex mathematical proofs and calculations - Advanced coding problems and algorithm design - Multi-step analytical reasoning tasks - Problems requiring diverse domain expertise - Tasks where single models show limitations The MoA approach uses a fixed 2-layer architecture: 1. Layer 1: Multiple reference models generate diverse responses in parallel (temp=0.6) 2. Layer 2: Aggregator model synthesizes the best elements into final response (temp=0.4) Args: user_prompt (str): The complex query or problem to solve reference_models (Optional[List[str]]): Custom reference models to use aggregator_model (Optional[str]): Custom aggregator model to use Returns: str: JSON string containing the MoA results with the following structure: { "success": bool, "response": str, "models_used": { "reference_models": List[str], "aggregator_model": str }, "processing_time": float } Raises: Exception: If MoA processing fails or API key is not set Nz...)r#rYrZreference_temperatureaggregator_temperaturemin_successful_referencesFr) parametersrMsuccessreference_responses_countfailed_models_count failed_modelsfinal_response_lengthprocessing_time_seconds models_usedz(Starting Mixture-of-Agents processing...z Query: %sdOPENROUTER_API_KEYz/OPENROUTER_API_KEY environment variable not setz5Using %s reference models in 2-layer MoA architecturez*Layer 1: Generating reference responses...c<g|]}t|tSr)rTREFERENCE_TEMPERATURE)rr"r#s rrz*mixture_of_agents_tool..9s6/ / /  &e[:O P P/ / / rz1Reference model results: %s successful, %s failedzFailed models: %s, z*Insufficient successful reference models (/z). Need at least z successful responses.rbrcrdz'Layer 2: Synthesizing final response...z(MoA processing completed in %.2f secondsT)rYrZ)rarrgrarerfrgmixture_of_agents_toolr5)indent ensure_asciizError in MoA processing: r:r;zMMoA processing failed. Please try again or use a single model for this query.)rarrgrMrM)datetimenowrJREFERENCE_MODELSrWrkAGGREGATOR_TEMPERATUREMIN_SUCCESSFUL_REFERENCESr>r?osgetenv ValueErrorrGgatherappendrFrr AGGREGATOR_SYSTEM_PROMPTrX total_seconds_debuglog_callsavejsondumpsrKrLrM)r#rYrZ start_timedebug_call_data ref_models agg_model model_resultssuccessful_responsesrd model_namer+rasuccessful_count failed_countaggregator_system_promptfinal_responseend_timeprocessing_timeresultrPrSs` rrnrnsR"&&((J9??? KTcT!2333y-.. PNOO O&9)9 $8(8  KSQ[__]]]  @AAA%n/ / / / #/ / /        " ,9 1 1 (J 1$++G4444$$Z0000344=))  GIY[ghhh  J NN. -0H0H I I I 7 7 7qJZqq]`ak]l]lqq@Yqqqrr r7G341=-.+8(  =>>>#? $ $ $  5 $  "        $((**#j0??AA >PPP&$.$-  &* "36~3F3F/05D12)/ )> & 0/BBB z&???? @@@8A88  T9t 444$((**#j0??AAg$4$H8H$4$H8H  $- 5D120/BBB z&?????????1@s 3KM P+CP& P+&P+ctS)z Check if all requirements for MoA tools are met. Returns: bool: True if requirements are met, False otherwise )check_openrouter_api_keyrrrcheck_moa_requirementsrs $ % %%rc tttttt tt ttz dt tddS)z Get the current MoA configuration settings. Returns: Dict[str, Any]: Dictionary containing all configuration parameters rmz models can fail)rYrZr]r^r_total_reference_modelsfailure_tolerance)rsrWrkrtrurJrrrget_moa_configurationrs[-,!6"8%>"%&6"7"7 #$4 5 58Q Q{{TWXhTiTi{{{  r__main__u"🤖 Mixture-of-Agents Tool Modulez2==================================================u3❌ OPENROUTER_API_KEY environment variable not setzBPlease set your API key: export OPENROUTER_API_KEY='your-key-here'z&Get API key at: https://openrouter.ai/u✅ OpenRouter API key foundu!🛠️ MoA tools ready for use!u ⚙️ Current Configuration:u 🤖 Reference models (z): rlu 🧠 Aggregator model: u" 🌡️ Reference temperature: r]u# 🌡️ Aggregator temperature: r^u 🛡️ Failure tolerance: ru" 📊 Minimum successful models: r_u' 🐛 Debug mode ENABLED - Session ID: z7 Debug logs will be saved to: ./logs/moa_tools_debug_z.jsonu> 🐛 Debug mode disabled (set MOA_TOOLS_DEBUG=true to enable)z Basic usage:z; from mixture_of_agents_tool import mixture_of_agents_toolz import asyncioz async def main():z, result = await mixture_of_agents_tool(z@ user_prompt='Solve this complex mathematical proof...'z )z print(result)z asyncio.run(main())z Best use cases:z0 - Complex mathematical proofs and calculationsz1 - Advanced coding problems and algorithm designz) - Multi-step analytical reasoning tasksz/ - Problems requiring diverse domain expertisez. - Tasks where single models show limitationsz Performance characteristics:z. - Higher latency due to multiple model callsz4 - Significantly improved quality for complex tasksz& - Parallel processing for efficiencyz - Optimized temperatures: z for reference models, z for aggregationz; - Token-efficient: only returns final aggregated responsez4 - Resilient: continues with partial model failureszC - Configurable: easy to modify models and settings at top of filez6 - State-of-the-art results on challenging benchmarksz Debug mode:z # Enable debug loggingz export MOA_TOOLS_DEBUG=truez; # Debug logs capture all MoA processing steps and metricsz3 # Logs saved to: ./logs/moa_tools_debug_UUID.json)registrymixture_of_agentsuIRoute a hard problem through multiple frontier LLMs collaboratively. Makes 5 API calls (4 reference models + 1 aggregator) with maximum reasoning effort — use sparingly for genuinely difficult problems. Best for: complex math, advanced algorithms, multi-step analytical reasoning, problems benefiting from diverse perspectives.objectstringzThe complex query or problem to solve using multiple AI models. Should be a challenging problem that benefits from diverse perspectives and collaborative reasoning.)type description)r propertiesrequired)namerr`moac Jt|ddS)Nr#r)r#)rnget)argskws rrs!5$((=Z\B]B]^^^rriTu🧠)rtoolsetschemahandlercheck_fn requires_envis_asyncemoji)NN)5__doc__rloggingrvrGrqtypingrrrrtools.openrouter_clientrrBrragent.auxiliary_clientr tools.debug_helpersr getLogger__name__r>rsrWrkrtrur{r}rLr floatinttupleboolrTrXrnrrprint api_availableexitconfigrJractive session_idtools.registryr MOA_SCHEMAregisterrrrrsS,,\  ,,,,,,,,,,,,yyyyyyyy??????,,,,,,  8 $ $ / k+< = = = 1 1S 1c 1 1 1 1$/ I/I/ I/I/I/ I/  I/  3T> I/I/I/I/^0 44444 4  4444r-1&*m@m@m@tCy)m@smm@ m@m@m@m@`&&&&&tCH~$ z E ./// E(OOO-,..M . CDDD RSSS 6777 Q ,--- E -...# " $ $F E ,--- E qcc&1C*D&E&E q q$))TZ[mTnJoJo q qrrr E Bf-?&@ B BCCC E Pv6M/N P PQQQ E R7O0P R RSSS E H62E+F H HIII E Tv6Q/R T TUUU}Q L9JLLMMM `HY```aaaa OPPP E  E GHHH E  E"III E     E 8999 E LMMM E) E     E !""" E  E <=== E =>>> E 5666 E ;<<< E :;;; E *+++ E :;;; E @AAA E 2333 E )>  Wm   AAA E GHHH E @AAA E OPPP E BCCC E/ E $%%% E )*** E GHHH E ?@@@ $##### _   F #O         ^ ^ #&'        r