fix: Cost calculation uses hardcoded Sonnet pricing - doesn't support Opus/Haiku models

[sagarsrc]

Scope

backend

What's broken?

The cost calculation in token_metrics.py uses hardcoded Claude Sonnet 4 pricing constants, but Claude Code can use different models (Opus 4.5, Opus 4, Haiku, etc.). This causes significant cost estimation errors.

File: packages/server/src/quickcall_supertrace/metrics/token_metrics.py:16-20

# Cost per 1M tokens (Claude Sonnet 4 pricing)
INPUT_COST_PER_M = 3.00
OUTPUT_COST_PER_M = 15.00
CACHE_READ_COST_PER_M = 0.30
CACHE_WRITE_COST_PER_M = 3.75

The model field is already being parsed in parser.py:211 but is not used in cost calculation.

Impact:

Model Used	Actual Cost	QuickCall Shows	Error
Opus 4.5	~$0.05	~$0.03	-40% underestimated
Opus 4/4.1	~$0.15	~$0.03	-80% underestimated
Haiku 4.5	~$0.01	~$0.03	+200% overestimated
Sonnet 4/4.5	~$0.03	~$0.03	✓ Correct

Expected behavior

Cost calculation should use model-specific pricing based on the model field from the API response.

Official Claude Pricing (per MTok):

Model	Input	Output	Cache Read	Cache Write (5m)
Opus 4.5	$5.00	$25.00	$0.50	$6.25
Opus 4/4.1	$15.00	$75.00	$1.50	$18.75
Sonnet 4/4.5	$3.00	$15.00	$0.30	$3.75
Haiku 4.5	$1.00	$5.00	$0.10	$1.25

Source: https://platform.claude.com/docs/en/about-claude/pricing

Version

0.2.8