# LLM Token Pricing Configuration
# Prices are per million tokens (as of Feb 2026)
# Update these values as pricing changes

models:
  # =============================================================================
  # Anthropic
  # =============================================================================
  claude-haiku-4-5:
    provider: anthropic
    input: 1.00
    output: 5.00
    cached_input: 0.10

  claude-sonnet-4:
    provider: anthropic
    input: 3.00
    output: 15.00
    cached_input: 0.30

  claude-opus-4:
    provider: anthropic
    input: 15.00
    output: 75.00
    cached_input: 1.50

  claude-sonnet-4-6:
    provider: anthropic
    input: 3.00
    output: 15.00
    cached_input: 0.30

  claude-opus-4-6:
    provider: anthropic
    input: 5.00
    output: 25.00
    cached_input: 0.50

  claude-opus-4-5:
    provider: anthropic
    input: 5.00
    output: 25.00
    cached_input: 0.50

  claude-sonnet-4-5:
    provider: anthropic
    input: 3.00
    output: 15.00
    cached_input: 0.30

  # Legacy models
  claude-3-5-sonnet-20241022:
    provider: anthropic
    input: 3.00
    output: 15.00
    cached_input: 0.30

  claude-3-5-haiku-20241022:
    provider: anthropic
    input: 0.80
    output: 4.00
    cached_input: 0.08

  # =============================================================================
  # OpenAI
  # =============================================================================

  # GPT-5 series
  gpt-5.2:
    provider: openai
    input: 1.75
    output: 14.00
    cached_input: 0.175

  gpt-5.2-pro:
    provider: openai
    input: 21.00
    output: 168.00

  gpt-5-mini:
    provider: openai
    input: 0.25
    output: 2.00
    cached_input: 0.025

  gpt-5-nano:
    provider: openai
    input: 0.05
    output: 0.40
    cached_input: 0.005

  # GPT-4.1 series
  gpt-4.1:
    provider: openai
    input: 2.00
    output: 8.00
    cached_input: 0.50

  gpt-4.1-mini:
    provider: openai
    input: 0.40
    output: 1.60
    cached_input: 0.10

  gpt-4.1-nano:
    provider: openai
    input: 0.10
    output: 0.40
    cached_input: 0.025

  # GPT-4o series (legacy)
  gpt-4o:
    provider: openai
    input: 2.50
    output: 10.00
    cached_input: 1.25

  gpt-4o-mini:
    provider: openai
    input: 0.15
    output: 0.60
    cached_input: 0.075

  gpt-4-turbo:
    provider: openai
    input: 10.00
    output: 30.00

  gpt-4:
    provider: openai
    input: 30.00
    output: 60.00

  gpt-3.5-turbo:
    provider: openai
    input: 0.50
    output: 1.50

  # Reasoning models (o-series)
  o1:
    provider: openai
    input: 15.00
    output: 60.00
    cached_input: 7.50
    reasoning: 60.00

  o1-mini:
    provider: openai
    input: 1.10
    output: 4.40
    cached_input: 0.55
    reasoning: 4.40

  o3-mini:
    provider: openai
    input: 1.10
    output: 4.40
    cached_input: 0.55
    reasoning: 4.40

  o4-mini:
    provider: openai
    input: 1.10
    output: 4.40
    cached_input: 0.275

  # =============================================================================
  # Google
  # =============================================================================
  gemini-1.5-pro:
    provider: google
    input: 1.25
    output: 5.00
    cached_input: 0.3125

  gemini-1.5-flash:
    provider: google
    input: 0.075
    output: 0.30
    cached_input: 0.01875

  gemini-2.0-flash:
    provider: google
    input: 0.10
    output: 0.40

  # =============================================================================
  # Perplexity (usage-based, approximate)
  # =============================================================================
  llama-3.1-sonar-small-128k-online:
    provider: perplexity
    input: 0.20
    output: 0.20

  llama-3.1-sonar-large-128k-online:
    provider: perplexity
    input: 1.00
    output: 1.00

  llama-3.1-sonar-huge-128k-online:
    provider: perplexity
    input: 5.00
    output: 5.00

# =============================================================================
# API Services - Extensible pricing for non-LLM APIs
# =============================================================================
#
# Pricing types:
#   - token: Price per million tokens (usage extracted from response)
#   - unit: Price per API unit (usage extracted or calculated from response)
#   - request: Fixed price per request
#
# To add a new provider:
#   1. Add entry with url_pattern to match HTTP calls
#   2. Set type and pricing
#   3. Optional: usage_path to extract usage from response body
#
services:
  # ---------------------------------------------------------------------------
  # Jina Reader API - Token-based
  # Pricing: https://jina.ai/reader/
  # ---------------------------------------------------------------------------
  jina:
    type: token
    url_pattern: "r.jina.ai"
    usage_path: "body.data.usage.tokens"
    per_million: 0.045                      # $0.045 per 1M tokens

  # ---------------------------------------------------------------------------
  # Semrush API - Unit-based (converted to dollars)
  # Pricing: https://developer.semrush.com/api/basics/api-units-balance/
  # ---------------------------------------------------------------------------
  semrush:
    type: unit
    url_pattern: "api.semrush.com"
    price_per_unit: 0.00025                 # ~$0.25 per 1000 units
    endpoints:
      backlinks_overview:
        pattern: "type=backlinks_overview"
        units_per_request: 10
      domain_organic:
        pattern: "type=domain_organic"
        units_per_line: 10
      phrase_related:
        pattern: "type=phrase_related"
        units_per_line: 40
      domain_ranks:
        pattern: "type=domain_ranks"
        units_per_request: 10

  # ---------------------------------------------------------------------------
  # Exa Research API - Cost from response
  # Pricing: https://exa.ai/pricing
  #
  # Exa's pricing structure (per 1k tasks):
  #   - Agent search operations: $5 (all models)
  #   - Agent page reads: $5 (exa-research/fast) or $10 (exa-research-pro) per 1k pages
  #   - Reasoning tokens: $5 per 1M tokens
  #
  # The API returns actual cost in response.body.costDollars.total
  # which accounts for searches, page reads, and reasoning tokens.
  # ---------------------------------------------------------------------------
  exa:
    type: response_cost
    url_pattern: "api.exa.ai"
    cost_path: "output.body.costDollars.total"
    # Exa uses async flow: POST creates task, GET returns results with costDollars
    # Only count GET responses that have costDollars.total - no fallback needed
    # For traces without verbose HTTP bodies, Exa costs will show as $0

  # ---------------------------------------------------------------------------
  # Tavily API - Request-based
  # Pricing: https://tavily.com/pricing
  # ---------------------------------------------------------------------------
  tavily:
    type: request
    url_pattern: "api.tavily.com"
    endpoints:
      search:
        pattern: "/search"
        price: 0.01
      extract:
        pattern: "/extract"
        price_per_item: 0.005
        items_path: "body.urls"

  # ---------------------------------------------------------------------------
  # DataForSEO API - Request-based
  # Pricing: https://dataforseo.com/pricing
  # ---------------------------------------------------------------------------
  dataforseo:
    type: request
    url_pattern: "api.dataforseo.com"
    endpoints:
      whois:
        pattern: "/whois"
        price: 0.015
      bulk_traffic:
        pattern: "/bulk_traffic"
        price_per_item: 0.075
        items_path: "body[0].targets"

  # ---------------------------------------------------------------------------
  # Perplexity API (direct HTTP calls, not via LLM SDK)
  # Pricing: https://docs.perplexity.ai/guides/pricing
  # ---------------------------------------------------------------------------
  perplexity_api:
    type: token
    url_pattern: "api.perplexity.ai"
    usage_path: "body.usage"
    input_field: "prompt_tokens"
    output_field: "completion_tokens"
    input_per_million: 1.00
    output_per_million: 1.00