Spaces:

codey-lab
/

Multi-LLM-API-Gateway

Running

File size: 21,063 Bytes

# =============================================================================
# .pyfun — PyFundaments App Configuration
# Single source of truth for app/* modules (provider, models, tools, hub)
# Part of: Universal MCP Hub on PyFundaments
# =============================================================================
# RULES:
#   - All values in double quotes "value"
#   - NO secrets here! Keys stay in .env → only ENV-VAR NAMES referenced here
#   - Comment-out unused sections with #  → keep structure, parsers need it!
#   - DO NOT DELETE headers or [X_END]   → parsers rely on these markers
#   - Empty/unused values: ""            → never leave bare =
# =============================================================================
# TIERS:
#   LAZY:       fill [HUB] + one [LLM_PROVIDER.*] only        → works
#   NORMAL:     + [SEARCH_PROVIDER.*] + [MODELS.*]             → works better
#   PRODUCTIVE: + [TOOLS] + [FALLBACK] + [HUB_LIMITS]          → full power
# =============================================================================
# DO NOT DELETE — file identifier used by all parsers
[PYFUN_FILE = .pyfun]

# =============================================================================
# HUB — Core identity & transport config
# =============================================================================
# =============================================================================
# HUB — Core identity & transport config
# =============================================================================
[HUB]
HUB_NAME            = "Multi-LLM-API-Gateway"
HUB_VERSION         = "1.0.1"
HUB_DESCRIPTION     = "Universal Hub built on PyFundaments"

# Transport: sse (legacy/HF Spaces fallback) | streamable-http (MCP spec 2025-11-25)
HUB_TRANSPORT       = "streamable-http"

# Stateless: true = HF Spaces + horizontal scaling safe
# false = nur wenn du Server-initiated Notifications brauchst
HUB_STATELESS       = "true"

HUB_HOST            = "0.0.0.0"
HUB_PORT            = "7860"
HUB_MODE            = "mcp"
HUB_SPACE_URL       = ""
[HUB_END]

# =============================================================================
# HUB_LIMITS — Request & retry behavior
# =============================================================================
[HUB_LIMITS]
MAX_PARALLEL_REQUESTS   = "5"
RETRY_COUNT             = "3"
RETRY_DELAY_SEC         = "2"
REQUEST_TIMEOUT_SEC     = "60"
SEARCH_TIMEOUT_SEC      = "30"
[HUB_LIMITS_END]

# =============================================================================
# PROVIDERS — All external API providers
# Secrets stay in .env! Only ENV-VAR NAMES are referenced here.
# =============================================================================
[PROVIDERS]

# ── LLM Providers ─────────────────────────────────────────────────────────────
[LLM_PROVIDERS]


  # later for customs assi
  [LLM_PROVIDER.smollm]
  active        = "true"
  base_url      = "https://codey-lab-smollm2-customs.hf.space/v1"
  env_key       = "SMOLLM_API_KEY"
  default_model = "smollm2-360m"
  models        = "smollm2-360m, codey-lab/model.universal-mcp-hub"
  fallback_to   = "gemini"
  [LLM_PROVIDER.smollm_END]

  [LLM_PROVIDER.anthropic]
  active              = "true"
  base_url            = "https://api.anthropic.com/v1"
  env_key             = "ANTHROPIC_API_KEY"
  api_version_header  = "2023-06-01"
  default_model       = "claude-haiku-4-5-20251001"
  models              = "claude-opus-4-6, claude-sonnet-4-6, claude-haiku-4-5-20251001"
  fallback_to         = "gemini"                  # ← gemini statt openrouter!
  [LLM_PROVIDER.anthropic_END]

  [LLM_PROVIDER.gemini]
  active        = "true"
  base_url      = "https://generativelanguage.googleapis.com/v1beta"
  env_key       = "GEMINI_API_KEY"
  default_model = "gemini-2.5-flash"
  models        = "gemini-2.0-flash, gemini-2.5-flash, gemini-2.5-flash-lite, gemini-3.1-flash-lite-preview, gemini-3-flash-preview, gemini-3.1-pro-preview, "
  fallback_to   = "openrouter"
  [LLM_PROVIDER.gemini_END]

  [LLM_PROVIDER.openrouter]
  active              = "true"
  base_url            = "https://openrouter.ai/api/v1"
  env_key             = "OPENROUTER_API_KEY"      # → .env: OPENROUTER_API_KEY=sk-or-...
  default_model       = "nvidia/nemotron-nano-9b-v2:free"
  models              = "meta-llama/llama-3-8b-instruct, mistralai/mistral-7b-instruct, nvidia/nemotron-nano-9b-v2:free, google/gemma-3-27b-it:free, openai/gpt-oss-20b:free, mistralai/Mistral-7B-Instruct-v0.3, meta-llama/Llama-3.3-70B-Instruct"
  fallback_to         = "" # last in chain, no further fallback                       
  [LLM_PROVIDER.openrouter_END]

  [LLM_PROVIDER.huggingface]
  active             = "false"
  base_url           = "https://api-inference.huggingface.co/v1"
  env_key            = "HF_TOKEN"                # → .env: HF_TOKEN=hf_...
  default_model = "meta-llama/Llama-3.1-8B-Instruct"
  models        = "meta-llama/Llama-3.1-8B-Instruct, mistralai/Mistral-Nemo-Instruct-2407, Qwen/Qwen2.5-72B-Instruct"
  fallback_to        = "" # last in chain, no further fallback
  [LLM_PROVIDER.huggingface_END]

  # ── Add more LLM providers below ──────────────────────────────────────────
  # [LLM_PROVIDER.mistral]
  # active            = "false"
  # base_url          = "https://api.mistral.ai/v1"
  # env_key           = "MISTRAL_API_KEY"
  # default_model     = "mistral-large-latest"
  # models            = "mistral-large-latest, mistral-small-latest"
  # fallback_to       = ""
  # [LLM_PROVIDER.mistral_END]

  # [LLM_PROVIDER.openai]
  # active            = "false"
  # base_url          = "https://api.openai.com/v1"
  # env_key           = "OPENAI_API_KEY"
  # default_model     = "gpt-4o"
  # models            = "gpt-4o, gpt-4o-mini, gpt-3.5-turbo"
  # fallback_to       = ""
  # [LLM_PROVIDER.openai_END]

[LLM_PROVIDERS_END]

# ── Search Providers ───────────────────────────────────────────────────────────
[SEARCH_PROVIDERS]

  [SEARCH_PROVIDER.brave]
  active              = "true"
  base_url            = "https://api.search.brave.com/res/v1/web/search"
  env_key             = "BRAVE_API_KEY"           # → .env: BRAVE_API_KEY=BSA...
  default_results     = "5"
  max_results         = "20"
  fallback_to         = "tavily"
  [SEARCH_PROVIDER.brave_END]

  [SEARCH_PROVIDER.tavily]
  active              = "true"
  base_url            = "https://api.tavily.com/search"
  env_key             = "TAVILY_API_KEY"          # → .env: TAVILY_API_KEY=tvly-...
  default_results     = "5"
  max_results         = "10"
  include_answer      = "true"                    # AI-synthesized answer
  fallback_to         = ""
  [SEARCH_PROVIDER.tavily_END]

  # ── Add more search providers below ───────────────────────────────────────
  # [SEARCH_PROVIDER.serper]
  # active            = "false"
  # base_url          = "https://google.serper.dev/search"
  # env_key           = "SERPER_API_KEY"
  # fallback_to       = ""
  # [SEARCH_PROVIDER.serper_END]

[SEARCH_PROVIDERS_END]

# ── Web / Action Providers (Webhooks, Bots, Social) ───────────────────────────
# [WEB_PROVIDERS]

  # [WEB_PROVIDER.discord]
  # active            = "false"
  # base_url          = "https://discord.com/api/v10"
  # env_key           = "BOT_TOKEN"
  # [WEB_PROVIDER.discord_END]

  # [WEB_PROVIDER.github]
  # active            = "false"
  # base_url          = "https://api.github.com"
  # env_key           = "GITHUB_TOKEN"
  # [WEB_PROVIDER.github_END]

# [WEB_PROVIDERS_END]

[PROVIDERS_END]

# =============================================================================
# MODELS — Token & rate limits per model
# Parser builds: MODELS[provider][model_name] → limits dict
# =============================================================================
[MODELS]

  [MODEL.claude-opus-4-6]
  provider            = "anthropic"
  context_tokens      = "200000"
  max_output_tokens   = "32000"
  requests_per_min    = "5"
  requests_per_day    = "300"
  cost_input_per_1k   = "0.015"                  # USD — update as pricing changes
  cost_output_per_1k  = "0.075"
  capabilities        = "text, code, analysis, vision"
  [MODEL.claude-opus-4-6_END]

  [MODEL.claude-sonnet-4-6]
  provider            = "anthropic"
  context_tokens      = "200000"
  max_output_tokens   = "16000"
  requests_per_min    = "50"
  requests_per_day    = "1000"
  cost_input_per_1k   = "0.003"
  cost_output_per_1k  = "0.015"
  capabilities        = "text, code, analysis, vision"
  [MODEL.claude-sonnet-4-6_END]

  [MODEL.claude-haiku-4-5-20251001]
  provider            = "anthropic"
  context_tokens      = "200000"
  max_output_tokens   = "8000"
  requests_per_min    = "50"
  requests_per_day    = "2000"
  cost_input_per_1k   = "0.00025"
  cost_output_per_1k  = "0.00125"
  capabilities        = "text, code, fast"
  [MODEL.claude-haiku-4-5-20251001_END]

  [MODEL.gemini-3.1-flash-lite-preview]
  provider            = "gemini"
  context_tokens      = "1000000"
  max_output_tokens   = "8192"
  requests_per_min    = "15"
  requests_per_day    = "1000"
  cost_input_per_1k   = "0.00010"
  cost_output_per_1k  = "0.00040"
  capabilities        = "text, code, vision, fast, cheap"
  [MODEL.gemini-3.1-flash-lite-preview_END]

  [MODEL.gemini-3-flash-preview]
  provider            = "gemini"
  context_tokens      = "1000000"
  max_output_tokens   = "8192"
  requests_per_min    = "10"
  requests_per_day    = "1000"
  cost_input_per_1k   = "0.00050"
  cost_output_per_1k  = "0.00300"
  capabilities        = "text, code, vision, audio, frontier"
  [MODEL.gemini-3-flash-preview_END]

  [MODEL.gemini-3.1-pro-preview]
  provider            = "gemini"
  context_tokens      = "1000000"
  max_output_tokens   = "64000"
  requests_per_min    = "5"
  requests_per_day    = "1000"
  cost_input_per_1k   = "0.00200"
  cost_output_per_1k  = "0.01200"
  capabilities        = "text, code, vision, audio, reasoning, agents"
  [MODEL.gemini-3.1-pro-preview_END]

  [MODEL.gemini-2.5-flash]
  provider            = "gemini"
  context_tokens      = "1000000"
  max_output_tokens   = "8192"
  requests_per_min    = "15"
  requests_per_day    = "1500"
  cost_input_per_1k   = "0.00010"
  cost_output_per_1k  = "0.00040"
  capabilities        = "text, code, vision, audio"
  [MODEL.gemini-2.5-flash_END]

  [MODEL.gemini-2.5-flash-lite]
  provider            = "gemini"
  context_tokens      = "1000000"
  max_output_tokens   = "8192"
  requests_per_min    = "15"
  requests_per_day    = "1500"
  cost_input_per_1k   = "0.00010"
  cost_output_per_1k  = "0.00040"
  capabilities        = "text, code, vision, audio"
  [MODEL.gemini-2.0-flash-lite_END]


  [MODEL.gemini-2.0-flash]
  provider            = "gemini"
  context_tokens      = "1000000"
  max_output_tokens   = "8192"
  requests_per_min    = "15"
  requests_per_day    = "1500"
  cost_input_per_1k   = "0.00010"
  cost_output_per_1k  = "0.00040"
  capabilities        = "text, code, vision, audio"
  [MODEL.gemini-2.0-flash_END]

  
  [MODEL.mistral-7b-instruct]
  provider            = "openrouter"
  context_tokens      = "32000"
  max_output_tokens   = "4096"
  requests_per_min    = "60"
  requests_per_day    = "10000"
  cost_input_per_1k   = "0.00006"
  cost_output_per_1k  = "0.00006"
  capabilities        = "text, code, fast, cheap"
  [MODEL.mistral-7b-instruct_END]
  
  [MODEL.dolphin-mistral-24b-venice-edition]
  provider            = "openrouter"
  context_tokens      = "32768"
  max_output_tokens   = "4096"
  requests_per_min    = ""
  requests_per_day    = ""
  cost_input_per_1k   = "0.00000"
  cost_output_per_1k  = "0.00000"
  capabilities        = "uncensored, text, code, fast, very cheap"
  [MODEL.dolphin-mistral-24b-venice-edition_END]

  [MODEL.nvidia-nemotron-nano-9b-v2]
  provider            = "openrouter"
  context_tokens      = "131072"
  max_output_tokens   = "4096"
  requests_per_min    = ""
  requests_per_day    = ""
  cost_input_per_1k   = "0.00000"
  cost_output_per_1k  = "0.00000"
  capabilities        = "text, code, reasoning, fast, free"
  [MODEL.nvidia-nemotron-nano-9b-v2_END]

  [MODEL.gemma-3-27b-it]
  provider            = "openrouter"
  context_tokens      = "131072"
  max_output_tokens   = "8192"
  requests_per_min    = ""
  requests_per_day    = ""
  cost_input_per_1k   = "0.00000"
  cost_output_per_1k  = "0.00000"
  capabilities        = "text, code, vision, multilingual, free"
  [MODEL.gemma-3-27b-it_END]

  [MODEL.gpt-oss-20b]
  provider            = "openrouter"
  context_tokens      = "128000"
  max_output_tokens   = "4096"
  requests_per_min    = ""
  requests_per_day    = ""
  cost_input_per_1k   = "0.00000"
  cost_output_per_1k  = "0.00000"
  capabilities        = "text, code, free"
  [MODEL.gpt-oss-20b_END]

  [MODEL.deephermes-3-llama-3-8b-preview]
  provider            = "openrouter"
  context_tokens      = "131072"
  max_output_tokens   = "4096"
  requests_per_min    = ""
  requests_per_day    = ""
  cost_input_per_1k   = "0.00000"
  cost_output_per_1k  = "0.00000"
  capabilities        = "text, code, reasoning, uncensored, free"
  [MODEL.deephermes-3-llama-3-8b-preview_END]

  [MODEL.Mistral-7B-Instruct-v0.3]
  provider            = "openrouter"
  context_tokens      = "32768"
  max_output_tokens   = "4096"
  requests_per_min    = "60"
  requests_per_day    = "10000"
  cost_input_per_1k   = "0.00006"
  cost_output_per_1k  = "0.00006"
  capabilities        = "text, code, fast, cheap"
  [MODEL.Mistral-7B-Instruct-v0.3_END]

  [MODEL.Llama-3.3-70B-Instruct]
  provider            = "openrouter"
  context_tokens      = "131072"
  max_output_tokens   = "8192"
  requests_per_min    = "60"
  requests_per_day    = "10000"
  cost_input_per_1k   = "0.00012"
  cost_output_per_1k  = "0.00030"
  capabilities        = "text, code, reasoning, multilingual"
  [MODEL.Llama-3.3-70B-Instruct_END]

  # customs llms
  [MODEL.smollm2-360m]
  provider           = "smollm"
  context_tokens     = "8192"
  max_output_tokens  = "300"
  requests_per_min   = "60"
  requests_per_day   = "10000"
  cost_input_per_1k  = "0.00000"
  cost_output_per_1k = "0.00000"
  capabilities       = "text, assistant, navigation, free, local"
  [MODEL.smollm2-360m_END]

  [MODEL.codey-lab/model.universal-mcp-hub]
  provider           = "smollm"
  context_tokens     = "8192"
  max_output_tokens  = "300"
  requests_per_min   = "60"
  requests_per_day   = "10000"
  cost_input_per_1k  = "0.00000"
  cost_output_per_1k = "0.00000"
  capabilities       = "text, assistant, navigation, free, custom, finetuned"
  [MODEL.codey-lab/model.universal-mcp-hub_END]

[MODELS_END]

# =============================================================================
# TOOLS — Tool definitions + provider mapping
# Tools are registered in mcp.py only if their provider ENV key exists!
# =============================================================================
[TOOLS]

  [TOOL.llm_complete]
  active              = "true"
  description         = "Send prompt to any configured LLM provider"
  provider_type       = "llm"
  default_provider    = "anthropic"
  timeout_sec         = "60"
  system_prompt       = "You are a helpful assistant integrated into the Universal MCP Hub. Answer concisely and only what is asked."
  [TOOL.llm_complete_END]

  [TOOL.code_review]
  active              = "true"
  description         = "Review code for bugs, security issues and improvements"
  provider_type       = "llm"
  default_provider    = "anthropic"
  timeout_sec         = "60"
  system_prompt       = "You are an expert code reviewer. Analyze the given code for bugs, security issues, and improvements. Be specific and concise."
  [TOOL.code_review_END]

  [TOOL.summarize]
  active              = "true"
  description         = "Summarize text into concise bullet points"
  provider_type       = "llm"
  default_provider    = "gemini"
  timeout_sec         = "30"
  system_prompt       = "You are a summarization expert. Summarize the given text in 3-5 bullet points. Be concise and accurate."
  [TOOL.summarize_END]

  [TOOL.translate]
  active              = "true"
  description         = "Translate text — auto-detects source language"
  provider_type       = "llm"
  default_provider    = "gemini"
  timeout_sec         = "30"
  system_prompt       = "You are a professional translator. Translate the given text accurately. Auto-detect the source language. Output only the translation, no explanation."
  [TOOL.translate_END]

  [TOOL.web_search]
  active              = "true"
  description         = "Search the web via configured search provider"
  provider_type       = "search"
  default_provider    = "brave"
  timeout_sec         = "30"
  [TOOL.web_search_END]

  [TOOL.db_query]
  active              = "true"
  description         = "Execute SELECT queries on connected database (read-only)"
  provider_type       = "db"
  readonly            = "true"
  timeout_sec         = "10"
  [TOOL.db_query_END]

  [TOOL.persist_result]
  active           = "true"
  description      = "Persist hub_state data to PostgreSQL for long-term storage"
  provider_type    = "persist"
  state_read_key   = ""          # welcher hub_state Key → leer = prompt direkt
  target_table     = "hub_results"
  timeout_sec      = "10"
  [TOOL.persist_result_END]

  # ── Future tools ──────────────────────────────────────────────────────────
  # [TOOL.image_gen]
  # active            = "false"
  # description       = "Generate images via configured provider"
  # provider_type     = "image"
  # default_provider  = ""
  # timeout_sec       = "120"
  # [TOOL.image_gen_END]

  # [TOOL.code_exec]
  # active            = "false"
  # description       = "Execute sandboxed code snippets"
  # provider_type     = "sandbox"
  # timeout_sec       = "30"
  # [TOOL.code_exec_END]

    # ── Shellmaster 2.0 ──────────────────────────────────────────────────────────

  [TOOL.shellmaster]
  active                     = "false"
  shellmaster_agent_url      = "http://localhost:5004"
  description                = "Generate safe shell commands for requested tasks"
  provider_type              = "llm"
  default_provider           = "smollm"
  timeout_sec                = "30"
  shellmaster_commands_file  = "shellmaster_commands.jsonl"
  shellmaster_commands_dataset_url = "" 
  shellmaster_customs_model_url  = ""
  system_prompt              = "You are ShellMaster. Generate safe shell commands.
                              ALWAYS include a backup and recovery plan.
                              Output JSON: {command, backup, recovery, risk}"
  [TOOL.shellmaster_END]

[TOOLS_END]

# =============================================================================
# DB_SYNC — Internal SQLite config for app/* IPC
# This is NOT the cloud DB — that lives in .env → DATABASE_URL
# =============================================================================
[DB_SYNC]
SQLITE_PATH         = "app/.hub_state.db"       # internal state, never commit!
SYNC_INTERVAL_SEC   = "30"                       # how often to flush to SQLite
MAX_CACHE_ENTRIES   = "1000"
[DB_SYNC_END]

# =============================================================================
# DEBUG — app/* debug behavior (fundaments debug stays in .env)
# =============================================================================
[DEBUG]
DEBUG               = "ON"                       # ON | OFF
DEBUG_LEVEL         = "FULL"                     # FULL | WARN | ERROR
LOG_FILE            = "hub_debug.log"
LOG_REQUESTS        = "true"                     # log every provider request
LOG_RESPONSES       = "false"                    # careful: may log sensitive data!
[DEBUG_END]

# =============================================================================
# PARSER_PROVIDER FREE ENDPOINTS — soon
# =============================================================================
[PARSER_PROVIDER]
#[PARSER_PROVIDER.metaculus]
#active    = "true"
#base_url  = "https://www.metaculus.com/api2"
#env_key   = ""          # public API, kein key!
#category  = "research"
#legal_de  = "true"      # ← Deutschland-Check direkt in config!
#[PARSER_PROVIDER.metaculus_END]

#[PARSER_PROVIDER.polymarket]
#active    = "false"
#base_url  = "https://gamma-api.polymarket.com"
#env_key   = ""
#legal_de  = "false"
#[PARSER_PROVIDER.polymarket_END]
[PARSER_PROVIDER_END]