mirrors/hermes-agent
1
0
Fork 0
mirror of https://github.com/NousResearch/hermes-agent.git synced 2026-07-20 15:33:54 +00:00
Code Issues Projects Releases Packages Wiki Activity Actions 7

feat(pricing): refresh Fireworks snapshot to 2026-07, cover full serverless catalog + cached picker pricing

Browse source 
- Refresh _OFFICIAL_DOCS_PRICING fireworks entries against current
  docs.fireworks.ai/serverless/pricing: qwen3p6-plus is gone (replaced
  by qwen3p7-plus); add glm-5p2/5p1, kimi-k2p7-code, deepseek-v4-flash,
  minimax-m3/m2p7, gpt-oss-120b/20b, and the routers/*-fast tiers with
  their distinct higher rates.
- Picker pricing via get_pricing_for_provider('fireworks'): pure dict
  transform over the shared models.dev in-memory/disk cache (1h TTL) +
  _pricing_cache memoization — no new network call on the picker path.
- Wire pricing display into the generic api-key-provider setup flow so
  Fireworks model pickers show $/M columns like OpenRouter/Nous do.
- Invariant tests: plugin fallback_models all priced, fast tiers price
  higher than standard, every row carries cache_read < input.
This commit is contained in:
Teknium 2026-07-16 00:58:48 -07:00
parent 365620ab28
commit fe5c0cb6c3
4 changed files with 274 additions and 13 deletions
Download patch file Download diff file Expand all files Collapse all files

162
agent/usage_pricing.py
View file

@ -613,6 +613,7 @@ _OFFICIAL_DOCS_PRICING: Dict[tuple[str, str], PricingEntry] = {
# through when configured with provider="fireworks". Fireworks publishes a
# cached_input rate per model alongside input/output, which maps to
# cache_read_cost_per_million. No separately published cache_write rate.
# Snapshot of https://docs.fireworks.ai/serverless/pricing (Standard tier).
(
"fireworks",
"kimi-k2p6",
@ -622,7 +623,29 @@ _OFFICIAL_DOCS_PRICING: Dict[tuple[str, str], PricingEntry] = {
cache_read_cost_per_million=Decimal("0.16"),
source="official_docs_snapshot",
source_url="https://docs.fireworks.ai/serverless/pricing",
pricing_version="fireworks-pricing-2026-05",
pricing_version="fireworks-pricing-2026-07",
),
(
"fireworks",
"kimi-k2p7-code",
): PricingEntry(
input_cost_per_million=Decimal("0.95"),
output_cost_per_million=Decimal("4.00"),
cache_read_cost_per_million=Decimal("0.19"),
source="official_docs_snapshot",
source_url="https://docs.fireworks.ai/serverless/pricing",
pricing_version="fireworks-pricing-2026-07",
),
(
"fireworks",
"glm-5p2",
): PricingEntry(
input_cost_per_million=Decimal("1.40"),
output_cost_per_million=Decimal("4.40"),
cache_read_cost_per_million=Decimal("0.14"),
source="official_docs_snapshot",
source_url="https://docs.fireworks.ai/serverless/pricing",
pricing_version="fireworks-pricing-2026-07",
),
(
"fireworks",
@ -633,18 +656,141 @@ _OFFICIAL_DOCS_PRICING: Dict[tuple[str, str], PricingEntry] = {
cache_read_cost_per_million=Decimal("0.145"),
source="official_docs_snapshot",
source_url="https://docs.fireworks.ai/serverless/pricing",
pricing_version="fireworks-pricing-2026-05",
pricing_version="fireworks-pricing-2026-07",
),
(
"fireworks",
"qwen3p6-plus",
"deepseek-v4-flash",
): PricingEntry(
input_cost_per_million=Decimal("0.50"),
output_cost_per_million=Decimal("3.00"),
cache_read_cost_per_million=Decimal("0.10"),
input_cost_per_million=Decimal("0.14"),
output_cost_per_million=Decimal("0.28"),
cache_read_cost_per_million=Decimal("0.028"),
source="official_docs_snapshot",
source_url="https://fireworks.ai/models/fireworks/qwen3p6-plus",
pricing_version="fireworks-pricing-2026-05",
source_url="https://docs.fireworks.ai/serverless/pricing",
pricing_version="fireworks-pricing-2026-07",
),
(
"fireworks",
"qwen3p7-plus",
): PricingEntry(
input_cost_per_million=Decimal("0.40"),
output_cost_per_million=Decimal("1.60"),
cache_read_cost_per_million=Decimal("0.08"),
source="official_docs_snapshot",
source_url="https://docs.fireworks.ai/serverless/pricing",
pricing_version="fireworks-pricing-2026-07",
),
(
"fireworks",
"minimax-m3",
): PricingEntry(
input_cost_per_million=Decimal("0.30"),
output_cost_per_million=Decimal("1.20"),
cache_read_cost_per_million=Decimal("0.06"),
source="official_docs_snapshot",
source_url="https://docs.fireworks.ai/serverless/pricing",
pricing_version="fireworks-pricing-2026-07",
),
(
"fireworks",
"gpt-oss-120b",
): PricingEntry(
input_cost_per_million=Decimal("0.15"),
output_cost_per_million=Decimal("0.60"),
cache_read_cost_per_million=Decimal("0.015"),
source="official_docs_snapshot",
source_url="https://docs.fireworks.ai/serverless/pricing",
pricing_version="fireworks-pricing-2026-07",
),
(
"fireworks",
"gpt-oss-20b",
): PricingEntry(
input_cost_per_million=Decimal("0.07"),
output_cost_per_million=Decimal("0.30"),
cache_read_cost_per_million=Decimal("0.035"),
source="official_docs_snapshot",
source_url="https://docs.fireworks.ai/serverless/pricing",
pricing_version="fireworks-pricing-2026-07",
),
(
"fireworks",
"glm-5p1",
): PricingEntry(
input_cost_per_million=Decimal("1.40"),
output_cost_per_million=Decimal("4.40"),
cache_read_cost_per_million=Decimal("0.26"),
source="official_docs_snapshot",
source_url="https://docs.fireworks.ai/serverless/pricing",
pricing_version="fireworks-pricing-2026-07",
),
(
"fireworks",
"minimax-m2p7",
): PricingEntry(
input_cost_per_million=Decimal("0.30"),
output_cost_per_million=Decimal("1.20"),
cache_read_cost_per_million=Decimal("0.06"),
source="official_docs_snapshot",
source_url="https://docs.fireworks.ai/serverless/pricing",
pricing_version="fireworks-pricing-2026-07",
),
# Fast/turbo serving tiers — exposed as accounts/fireworks/routers/<name>,
# so rsplit("/", 1) yields these distinct ids with their own (higher) rates.
(
"fireworks",
"kimi-k2p6-fast",
): PricingEntry(
input_cost_per_million=Decimal("2.00"),
output_cost_per_million=Decimal("8.00"),
cache_read_cost_per_million=Decimal("0.30"),
source="official_docs_snapshot",
source_url="https://docs.fireworks.ai/serverless/pricing",
pricing_version="fireworks-pricing-2026-07",
),
(
"fireworks",
"kimi-k2p6-turbo",
): PricingEntry(
input_cost_per_million=Decimal("2.00"),
output_cost_per_million=Decimal("8.00"),
cache_read_cost_per_million=Decimal("0.30"),
source="official_docs_snapshot",
source_url="https://docs.fireworks.ai/serverless/pricing",
pricing_version="fireworks-pricing-2026-07",
),
(
"fireworks",
"kimi-k2p7-code-fast",
): PricingEntry(
input_cost_per_million=Decimal("1.90"),
output_cost_per_million=Decimal("8.00"),
cache_read_cost_per_million=Decimal("0.38"),
source="official_docs_snapshot",
source_url="https://docs.fireworks.ai/serverless/pricing",
pricing_version="fireworks-pricing-2026-07",
),
(
"fireworks",
"glm-5p2-fast",
): PricingEntry(
input_cost_per_million=Decimal("2.10"),
output_cost_per_million=Decimal("6.60"),
cache_read_cost_per_million=Decimal("0.21"),
source="official_docs_snapshot",
source_url="https://docs.fireworks.ai/serverless/pricing",
pricing_version="fireworks-pricing-2026-07",
),
(
"fireworks",
"glm-5p1-fast",
): PricingEntry(
input_cost_per_million=Decimal("2.80"),
output_cost_per_million=Decimal("8.80"),
cache_read_cost_per_million=Decimal("0.52"),
source="official_docs_snapshot",
source_url="https://docs.fireworks.ai/serverless/pricing",
pricing_version="fireworks-pricing-2026-07",
),
}

13
hermes_cli/model_setup_flows.py
View file

@ -2857,9 +2857,22 @@ def _model_flow_api_key_provider(config, provider_id, current_model=""):
model_list = list(dict.fromkeys(mid for mid in model_list if mid))
if model_list:
# Per-model pricing, when the provider supports it (fireworks via the
# models.dev disk cache, novita/deepinfra via their cached /models
# endpoints). get_pricing_for_provider() is memoized in-process and
# returns {} for providers without pricing — never a blocking fetch
# beyond the catalog lookup that already happened above.
pricing: dict = {}
try:
from hermes_cli.models import get_pricing_for_provider
pricing = get_pricing_for_provider(provider_id) or {}
except Exception:
pricing = {}
selected = _prompt_model_selection(
model_list,
current_model=current_model,
pricing=pricing,
confirm_provider=provider_id,
confirm_base_url=effective_base,
confirm_api_key=existing_key,

51
hermes_cli/models.py
View file

@ -1654,6 +1654,8 @@ def get_pricing_for_provider(provider: str, *, force_refresh: bool = False) -> d
return _fetch_novita_pricing(force_refresh=force_refresh)
if normalized == "deepinfra":
return _fetch_deepinfra_pricing(force_refresh=force_refresh)
if normalized == "fireworks":
return _fireworks_pricing_from_models_dev(force_refresh=force_refresh)
if normalized == "nous":
api_key, base_url = _resolve_nous_pricing_credentials()
if base_url:
@ -1670,6 +1672,55 @@ def get_pricing_for_provider(provider: str, *, force_refresh: bool = False) -> d
return {}
def _fireworks_pricing_from_models_dev(
*,
force_refresh: bool = False,
) -> dict[str, dict[str, str]]:
"""Derive Fireworks picker pricing from the models.dev registry cache.
No dedicated network fetch: ``fetch_models_dev()`` already maintains an
in-memory + disk cache (1h TTL) that every picker surface shares, so this
is a pure dict transform on the picker path — no added latency and no
per-render network call. Results are additionally memoized in
``_pricing_cache`` so repeated menu renders within a process are free.
models.dev publishes Fireworks costs in USD per 1M tokens; the shared
pricing formatter expects per-token strings, so divide by 1M.
"""
cache_key = "models.dev/fireworks"
if not force_refresh and cache_key in _pricing_cache:
return _pricing_cache[cache_key]
result: dict[str, dict[str, str]] = {}
try:
from agent.models_dev import _get_provider_models
models = _get_provider_models("fireworks") or {}
for mid, entry in models.items():
if not isinstance(entry, dict):
continue
cost = entry.get("cost")
if not isinstance(cost, dict):
continue
inp = cost.get("input")
out = cost.get("output")
if inp is None and out is None:
continue
row: dict[str, str] = {
"prompt": str(float(inp or 0) / 1_000_000),
"completion": str(float(out or 0) / 1_000_000),
}
cache_read = cost.get("cache_read")
if cache_read:
row["input_cache_read"] = str(float(cache_read) / 1_000_000)
result[str(mid)] = row
except Exception:
result = {}
_pricing_cache[cache_key] = result
return result
def _fetch_novita_pricing(
timeout: float = 8.0,
*,

61
tests/agent/test_usage_pricing.py
View file

@ -351,10 +351,10 @@ def test_fireworks_base_url_host_match_alone_routes_to_pricing():
assert float(entry.output_cost_per_million) == 3.48
def test_fireworks_qwen3p6_plus_estimate_usage_cost():
"""End-to-end: Fireworks Qwen3.6-Plus sessions report a dollar estimate."""
def test_fireworks_qwen3p7_plus_estimate_usage_cost():
"""End-to-end: Fireworks Qwen3.7-Plus sessions report a dollar estimate."""
result = estimate_usage_cost(
"accounts/fireworks/models/qwen3p6-plus",
"accounts/fireworks/models/qwen3p7-plus",
CanonicalUsage(input_tokens=1_000_000, output_tokens=500_000),
provider="fireworks",
base_url="https://api.fireworks.ai/inference/v1",
@ -362,5 +362,56 @@ def test_fireworks_qwen3p6_plus_estimate_usage_cost():
assert result.status == "estimated"
assert result.amount_usd is not None
# 1M input × $0.50/M + 500K output × $3.00/M = $0.50 + $1.50 = $2.00
assert float(result.amount_usd) == 2.00
# 1M input × $0.40/M + 500K output × $1.60/M = $0.40 + $0.80 = $1.20
assert float(result.amount_usd) == 1.20
def test_fireworks_router_fast_tier_prices_distinctly():
"""Fast serving tiers live under accounts/fireworks/routers/<name>-fast and
bill at higher rates than the standard model — the routing layer's
rsplit("/", 1) must land on the distinct fast-tier entry."""
standard = get_pricing_entry(
"accounts/fireworks/models/kimi-k2p6",
provider="fireworks",
base_url="https://api.fireworks.ai/inference/v1",
)
fast = get_pricing_entry(
"accounts/fireworks/routers/kimi-k2p6-fast",
provider="fireworks",
base_url="https://api.fireworks.ai/inference/v1",
)
assert standard is not None and fast is not None
assert fast.input_cost_per_million > standard.input_cost_per_million
assert fast.output_cost_per_million > standard.output_cost_per_million
def test_fireworks_plugin_fallback_models_all_have_pricing():
"""Invariant: every model in the Fireworks provider plugin's
fallback_models (the picker's curated safety net) must resolve to a
pricing entry — otherwise the default picker choices bill as unknown."""
from providers import get_provider_profile
profile = get_provider_profile("fireworks")
assert profile is not None
for mid in profile.fallback_models:
entry = get_pricing_entry(
mid,
provider="fireworks",
base_url="https://api.fireworks.ai/inference/v1",
)
assert entry is not None, f"no pricing entry for fallback model {mid}"
assert entry.input_cost_per_million is not None, mid
def test_fireworks_rows_all_carry_cache_read_pricing():
"""Invariant: Fireworks publishes cached-input rates for every serverless
model, and Hermes prompt caching is active on Fireworks sessions — every
snapshot row must carry a cache_read rate cheaper than fresh input."""
from agent.usage_pricing import _OFFICIAL_DOCS_PRICING
fw_rows = [k for k in _OFFICIAL_DOCS_PRICING if k[0] == "fireworks"]
assert fw_rows, "expected at least one fireworks pricing row"
for key in fw_rows:
entry = _OFFICIAL_DOCS_PRICING[key]
assert entry.cache_read_cost_per_million is not None, key
assert entry.cache_read_cost_per_million < entry.input_cost_per_million, key