spendgate 0.1.0__py3-none-any.whl
This diff represents the content of publicly available package versions that have been released to one of the supported registries. The information contained in this diff is provided for informational purposes only and reflects changes between package versions as they appear in their respective public registries.
- spendgate/__init__.py +1 -0
- spendgate/catalog/prices.yaml +342 -0
- spendgate/catalog/schema.json +135 -0
- spendgate/cli/__init__.py +5 -0
- spendgate/cli/client.py +58 -0
- spendgate/cli/commands.py +231 -0
- spendgate/cli/entry.py +185 -0
- spendgate/cli/render.py +26 -0
- spendgate/cli/runtime.py +52 -0
- spendgate/cli/serve.py +163 -0
- spendgate/client/__init__.py +41 -0
- spendgate/client/anthropic.py +64 -0
- spendgate/client/errors.py +207 -0
- spendgate/client/openai.py +64 -0
- spendgate/client/session.py +158 -0
- spendgate/config/__init__.py +17 -0
- spendgate/config/load.py +407 -0
- spendgate/config/schema.py +234 -0
- spendgate/config/secrets.py +63 -0
- spendgate/config/yamlio.py +37 -0
- spendgate/core/__init__.py +1 -0
- spendgate/core/audio.py +427 -0
- spendgate/core/budgets.py +194 -0
- spendgate/core/catalog.py +294 -0
- spendgate/core/cost.py +240 -0
- spendgate/core/errors.py +106 -0
- spendgate/core/estimate.py +388 -0
- spendgate/core/health.py +122 -0
- spendgate/core/limits.py +206 -0
- spendgate/core/money.py +191 -0
- spendgate/core/outcomes.py +166 -0
- spendgate/core/policy.py +199 -0
- spendgate/core/sessions.py +128 -0
- spendgate/gateway/__init__.py +1 -0
- spendgate/gateway/admin.py +126 -0
- spendgate/gateway/admin_budgets.py +172 -0
- spendgate/gateway/admin_codec.py +149 -0
- spendgate/gateway/admin_http.py +297 -0
- spendgate/gateway/admin_models.py +131 -0
- spendgate/gateway/admin_reports.py +166 -0
- spendgate/gateway/admin_scopes.py +215 -0
- spendgate/gateway/app.py +177 -0
- spendgate/gateway/bootstrap.py +133 -0
- spendgate/gateway/logs.py +92 -0
- spendgate/gateway/maintenance.py +57 -0
- spendgate/gateway/metrics.py +267 -0
- spendgate/gateway/observability.py +135 -0
- spendgate/gateway/wiring.py +163 -0
- spendgate/ledger/__init__.py +1 -0
- spendgate/ledger/admin_engine.py +87 -0
- spendgate/ledger/admin_flow.py +360 -0
- spendgate/ledger/admin_port.py +131 -0
- spendgate/ledger/admin_reads.py +229 -0
- spendgate/ledger/common.py +217 -0
- spendgate/ledger/engine.py +299 -0
- spendgate/ledger/hold_flow.py +298 -0
- spendgate/ledger/ids.py +12 -0
- spendgate/ledger/key_flow.py +318 -0
- spendgate/ledger/key_types.py +102 -0
- spendgate/ledger/keys.py +200 -0
- spendgate/ledger/migrations/__init__.py +34 -0
- spendgate/ledger/migrations/postgres/0001_init.sql +204 -0
- spendgate/ledger/migrations/postgres/0002_usage.sql +5 -0
- spendgate/ledger/migrations/postgres/0003_scope_display_name.sql +2 -0
- spendgate/ledger/migrations/postgres/0004_calls_http_status_check.sql +3 -0
- spendgate/ledger/migrations/sqlite/0001_init.sql +194 -0
- spendgate/ledger/migrations/sqlite/0002_usage.sql +5 -0
- spendgate/ledger/migrations/sqlite/0003_scope_display_name.sql +2 -0
- spendgate/ledger/migrations/sqlite/0004_calls_http_status_check.sql +12 -0
- spendgate/ledger/port.py +368 -0
- spendgate/ledger/postgres.py +158 -0
- spendgate/ledger/postgres_driver.py +109 -0
- spendgate/ledger/rows.py +83 -0
- spendgate/ledger/settle_flow.py +272 -0
- spendgate/ledger/sql.py +365 -0
- spendgate/ledger/sqlite.py +220 -0
- spendgate/ledger/usage.py +290 -0
- spendgate/ledger/usage_flow.py +104 -0
- spendgate/local/__init__.py +16 -0
- spendgate/local/bridge.py +107 -0
- spendgate/local/guard.py +280 -0
- spendgate/local/loop.py +44 -0
- spendgate/local/principal.py +71 -0
- spendgate/local/runtime.py +156 -0
- spendgate/local/sdk.py +55 -0
- spendgate/local/session.py +27 -0
- spendgate/local/setup.py +82 -0
- spendgate/local/transport.py +126 -0
- spendgate/local/transport_httpx2.py +100 -0
- spendgate/pipeline/__init__.py +1 -0
- spendgate/pipeline/adapters.py +37 -0
- spendgate/pipeline/answers.py +43 -0
- spendgate/pipeline/auth.py +40 -0
- spendgate/pipeline/closing.py +171 -0
- spendgate/pipeline/intake.py +248 -0
- spendgate/pipeline/leases.py +102 -0
- spendgate/pipeline/model.py +174 -0
- spendgate/pipeline/multipart.py +121 -0
- spendgate/pipeline/pipeline.py +408 -0
- spendgate/pipeline/replies.py +144 -0
- spendgate/pipeline/routing.py +435 -0
- spendgate/pipeline/settle.py +212 -0
- spendgate/pipeline/steps.py +347 -0
- spendgate/pipeline/streaming.py +335 -0
- spendgate/pipeline/transport.py +98 -0
- spendgate/pipeline/upkeep.py +46 -0
- spendgate/pipeline/watch.py +48 -0
- spendgate/providers/__init__.py +1 -0
- spendgate/providers/anthropic.py +381 -0
- spendgate/providers/azure_openai.py +146 -0
- spendgate/providers/azure_openai_auth.py +74 -0
- spendgate/providers/azure_speech.py +110 -0
- spendgate/providers/azure_speech_shapes.py +216 -0
- spendgate/providers/bedrock.py +231 -0
- spendgate/providers/bedrock_request.py +239 -0
- spendgate/providers/bedrock_stream.py +349 -0
- spendgate/providers/elevenlabs.py +196 -0
- spendgate/providers/google.py +326 -0
- spendgate/providers/google_shapes.py +121 -0
- spendgate/providers/google_stream.py +90 -0
- spendgate/providers/openai.py +76 -0
- spendgate/providers/openai_shapes.py +425 -0
- spendgate/providers/openai_usage.py +143 -0
- spendgate/providers/port.py +332 -0
- spendgate/providers/sse.py +169 -0
- spendgate/py.typed +0 -0
- spendgate-0.1.0.dist-info/METADATA +188 -0
- spendgate-0.1.0.dist-info/RECORD +131 -0
- spendgate-0.1.0.dist-info/WHEEL +4 -0
- spendgate-0.1.0.dist-info/entry_points.txt +3 -0
- spendgate-0.1.0.dist-info/licenses/LICENSE +202 -0
spendgate/__init__.py
ADDED
|
@@ -0,0 +1 @@
|
|
|
1
|
+
"""spendgate: a small, auditable LLM spend guard."""
|
|
@@ -0,0 +1,342 @@
|
|
|
1
|
+
# spendgate seed price catalog (schema: schema.json). Prices are quoted decimal strings in the entry's currency.
|
|
2
|
+
# Compiled 2026-11-13 from each provider's own pages / public price feeds (see source_url and per-entry notes).
|
|
3
|
+
# verified: true = every price on the entry was read today from the provider's own page or official public price feed.
|
|
4
|
+
# No Realtime entries (on the roadmap). Bedrock region used for every bedrock entry: us-east-1.
|
|
5
|
+
# Known schema gaps: Anthropic/Bedrock have 5m and 1h cache-write prices but only one cache_write meter exists, so the
|
|
6
|
+
# dearer 1h price is stored (never lower). Batch/Flex/Priority variants are not seeded. Vertex AI pricing page could not be
|
|
7
|
+
# read, so no Vertex-specific price is encoded (google/* entries are Gemini API prices).
|
|
8
|
+
catalog_version: 1
|
|
9
|
+
updated: "2026-11-13"
|
|
10
|
+
|
|
11
|
+
providers:
|
|
12
|
+
openai: { free_statuses: [429] }
|
|
13
|
+
azure_openai: { free_statuses: [429] }
|
|
14
|
+
anthropic: { free_statuses: [429, 529] }
|
|
15
|
+
bedrock: { free_statuses: [429] }
|
|
16
|
+
google: { free_statuses: [429] }
|
|
17
|
+
azure_speech: { free_statuses: [429] }
|
|
18
|
+
elevenlabs: { free_statuses: [429] }
|
|
19
|
+
|
|
20
|
+
models:
|
|
21
|
+
# ===================== OpenAI (direct) =====================
|
|
22
|
+
# Page: developers.openai.com/api/docs/pricing and /models/<id>. Long context: prompts over 272K input tokens price the
|
|
23
|
+
# whole request at 2x input/cache, 1.5x output. Max output 128,000, context 1,050,000 (model pages).
|
|
24
|
+
- id: openai/gpt-6-astra
|
|
25
|
+
mode: responses
|
|
26
|
+
currency: USD
|
|
27
|
+
effective_from: "2026-11-13"
|
|
28
|
+
source_url: https://developers.openai.com/api/docs/models/gpt-6-astra
|
|
29
|
+
verified: true
|
|
30
|
+
checked_on: "2026-11-13"
|
|
31
|
+
note: "Flagship. Standard tier. Pricing page lists 10/1/50; model page adds cache write 12.50."
|
|
32
|
+
deployment_type: standard
|
|
33
|
+
meters: { input_per_mtok: "10", cached_input_per_mtok: "1", cache_write_per_mtok: "12.5", output_per_mtok: "50" }
|
|
34
|
+
long_context: { threshold_input_tokens: 272000, input_multiplier: "2", output_multiplier: "1.5" }
|
|
35
|
+
limits: { max_output_tokens: 128000, context_window: 1050000 }
|
|
36
|
+
- id: openai/gpt-6-luna
|
|
37
|
+
mode: responses
|
|
38
|
+
currency: USD
|
|
39
|
+
effective_from: "2026-11-13"
|
|
40
|
+
source_url: https://developers.openai.com/api/docs/models/gpt-6-luna
|
|
41
|
+
verified: true
|
|
42
|
+
checked_on: "2026-11-13"
|
|
43
|
+
note: "Small/cheap model. Standard tier."
|
|
44
|
+
deployment_type: standard
|
|
45
|
+
meters: { input_per_mtok: "0.1", cached_input_per_mtok: "0.01", cache_write_per_mtok: "0.125", output_per_mtok: "0.5" }
|
|
46
|
+
long_context: { threshold_input_tokens: 272000, input_multiplier: "2", output_multiplier: "1.5" }
|
|
47
|
+
limits: { max_output_tokens: 128000, context_window: 1050000 }
|
|
48
|
+
- id: openai/text-embedding-3-small
|
|
49
|
+
mode: embedding
|
|
50
|
+
currency: USD
|
|
51
|
+
effective_from: "2026-11-13"
|
|
52
|
+
source_url: https://developers.openai.com/api/docs/pricing
|
|
53
|
+
verified: true
|
|
54
|
+
checked_on: "2026-11-13"
|
|
55
|
+
meters: { input_per_mtok: "0.02" }
|
|
56
|
+
|
|
57
|
+
# --- OpenAI speech-to-text (pricing page and /models/<id> pages read 2026-11-18). Not encoded: gpt-live-transcribe and
|
|
58
|
+
# gpt-realtime-whisper (Realtime only, on the roadmap); gpt-4o-transcribe, gpt-4o-mini-transcribe and gpt-4o-transcribe-diarize
|
|
59
|
+
# (deprecated, shutdown 2027-02-26; the pages disagree on whether their token prices are audio or text tokens). ---
|
|
60
|
+
- id: openai/gpt-transcribe
|
|
61
|
+
mode: transcription
|
|
62
|
+
currency: USD
|
|
63
|
+
effective_from: "2026-11-18"
|
|
64
|
+
source_url: https://developers.openai.com/api/docs/pricing
|
|
65
|
+
verified: true
|
|
66
|
+
checked_on: "2026-11-18"
|
|
67
|
+
note: "0.0045 USD per minute (pricing page and model page). No deprecation notice."
|
|
68
|
+
meters: { audio_input_per_minute: "0.0045" }
|
|
69
|
+
- id: openai/whisper-1
|
|
70
|
+
mode: transcription
|
|
71
|
+
currency: USD
|
|
72
|
+
effective_from: "2026-11-18"
|
|
73
|
+
source_url: https://developers.openai.com/api/docs/pricing
|
|
74
|
+
verified: true
|
|
75
|
+
checked_on: "2026-11-18"
|
|
76
|
+
note: "0.006 USD per minute (pricing page lists 'Whisper'). Deprecated 2026-08-26; shutdown 2027-02-26 (deprecations page)."
|
|
77
|
+
meters: { audio_input_per_minute: "0.006" }
|
|
78
|
+
deprecation_date: "2027-02-26"
|
|
79
|
+
|
|
80
|
+
# ===================== Azure OpenAI (Foundry Models) =====================
|
|
81
|
+
# Source: Azure Retail Prices API (https://prices.azure.com/api/retail/prices, serviceName 'Foundry Models'), read 2026-11-13.
|
|
82
|
+
# Stored meters are Global Standard; price_multiplier 1.20 = highest Data Zone uplift seen in any region (1.10 in US zones,
|
|
83
|
+
# 1.20 in EU). Long context (over 272K input) is 2x input/cache, 1.5x output, matching the LongCo meters. Cache write meter
|
|
84
|
+
# is published, so it is encoded. max_output_tokens 128000 follows OpenAI's model page (not re-checked on Azure docs).
|
|
85
|
+
- id: azure_openai/gpt-6-astra
|
|
86
|
+
mode: responses
|
|
87
|
+
currency: USD
|
|
88
|
+
effective_from: "2026-09-01"
|
|
89
|
+
source_url: https://prices.azure.com/api/retail/prices
|
|
90
|
+
verified: true
|
|
91
|
+
checked_on: "2026-11-13"
|
|
92
|
+
note: "Flagship. Global 10/1/12.5/50 x 1.20 Data Zone upper bound (retail meters DZ 11-12 / 55-60)."
|
|
93
|
+
deployment_type: data_zone
|
|
94
|
+
price_multiplier: "1.20"
|
|
95
|
+
meters: { input_per_mtok: "10", cached_input_per_mtok: "1", cache_write_per_mtok: "12.5", output_per_mtok: "50" }
|
|
96
|
+
long_context: { threshold_input_tokens: 272000, input_multiplier: "2", output_multiplier: "1.5" }
|
|
97
|
+
limits: { max_output_tokens: 128000, context_window: 1050000 }
|
|
98
|
+
- id: azure_openai/gpt-6-luna
|
|
99
|
+
mode: responses
|
|
100
|
+
currency: USD
|
|
101
|
+
effective_from: "2026-09-01"
|
|
102
|
+
source_url: https://prices.azure.com/api/retail/prices
|
|
103
|
+
verified: true
|
|
104
|
+
checked_on: "2026-11-13"
|
|
105
|
+
note: "Small/cheap. Global 0.10/0.01/0.125/0.50 x 1.20 Data Zone upper bound."
|
|
106
|
+
deployment_type: data_zone
|
|
107
|
+
price_multiplier: "1.20"
|
|
108
|
+
meters: { input_per_mtok: "0.1", cached_input_per_mtok: "0.01", cache_write_per_mtok: "0.125", output_per_mtok: "0.5" }
|
|
109
|
+
long_context: { threshold_input_tokens: 272000, input_multiplier: "2", output_multiplier: "1.5" }
|
|
110
|
+
limits: { max_output_tokens: 128000, context_window: 1050000 }
|
|
111
|
+
- id: azure_openai/gpt-6.1-terra
|
|
112
|
+
mode: responses
|
|
113
|
+
currency: USD
|
|
114
|
+
effective_from: "2026-10-01"
|
|
115
|
+
source_url: https://prices.azure.com/api/retail/prices
|
|
116
|
+
verified: true
|
|
117
|
+
checked_on: "2026-11-13"
|
|
118
|
+
note: "Global 1.00/0.05/1.25/5.00 x 1.20 Data Zone upper bound."
|
|
119
|
+
deployment_type: data_zone
|
|
120
|
+
price_multiplier: "1.20"
|
|
121
|
+
meters: { input_per_mtok: "1", cached_input_per_mtok: "0.05", cache_write_per_mtok: "1.25", output_per_mtok: "5" }
|
|
122
|
+
long_context: { threshold_input_tokens: 272000, input_multiplier: "2", output_multiplier: "1.5" }
|
|
123
|
+
limits: { max_output_tokens: 128000, context_window: 1050000 }
|
|
124
|
+
- id: azure_openai/gpt-6.1-sol
|
|
125
|
+
mode: responses
|
|
126
|
+
currency: USD
|
|
127
|
+
effective_from: "2026-10-01"
|
|
128
|
+
source_url: https://prices.azure.com/api/retail/prices
|
|
129
|
+
verified: true
|
|
130
|
+
checked_on: "2026-11-13"
|
|
131
|
+
note: "Global 2.00/0.10/2.50/10.00 x 1.20 Data Zone upper bound."
|
|
132
|
+
deployment_type: data_zone
|
|
133
|
+
price_multiplier: "1.20"
|
|
134
|
+
meters: { input_per_mtok: "2", cached_input_per_mtok: "0.1", cache_write_per_mtok: "2.5", output_per_mtok: "10" }
|
|
135
|
+
long_context: { threshold_input_tokens: 272000, input_multiplier: "2", output_multiplier: "1.5" }
|
|
136
|
+
limits: { max_output_tokens: 128000, context_window: 1050000 }
|
|
137
|
+
|
|
138
|
+
# --- Azure OpenAI speech-to-text (highest regional meter = upper bound; all from the Azure Retail Prices API) ---
|
|
139
|
+
- id: azure_openai/whisper
|
|
140
|
+
mode: transcription
|
|
141
|
+
currency: USD
|
|
142
|
+
effective_from: "2026-11-13"
|
|
143
|
+
source_url: https://prices.azure.com/api/retail/prices
|
|
144
|
+
verified: true
|
|
145
|
+
checked_on: "2026-11-13"
|
|
146
|
+
note: "Highest regional batch meter 0.4752 USD/h (global 0.36). Retires 2026-12-15 per learn.microsoft.com/azure/foundry/openai/concepts/model-retirement-schedule, read 2026-11-18."
|
|
147
|
+
meters: { audio_input_per_minute: "0.00792" }
|
|
148
|
+
deprecation_date: "2026-12-15"
|
|
149
|
+
- id: azure_openai/gpt-transcribe
|
|
150
|
+
mode: transcription
|
|
151
|
+
currency: USD
|
|
152
|
+
effective_from: "2026-09-01"
|
|
153
|
+
source_url: https://prices.azure.com/api/retail/prices
|
|
154
|
+
verified: true
|
|
155
|
+
checked_on: "2026-11-13"
|
|
156
|
+
note: "Data Zone meter 0.297 USD/h (global 0.27)."
|
|
157
|
+
meters: { audio_input_per_minute: "0.00495" }
|
|
158
|
+
- id: azure_openai/gpt-live-transcribe
|
|
159
|
+
mode: transcription
|
|
160
|
+
currency: USD
|
|
161
|
+
effective_from: "2026-09-01"
|
|
162
|
+
source_url: https://prices.azure.com/api/retail/prices
|
|
163
|
+
verified: true
|
|
164
|
+
checked_on: "2026-11-13"
|
|
165
|
+
note: "Data Zone meter 1.122 USD/h (global 1.02)."
|
|
166
|
+
meters: { audio_input_per_minute: "0.0187" }
|
|
167
|
+
- id: azure_openai/gpt-4o-transcribe
|
|
168
|
+
mode: transcription
|
|
169
|
+
currency: USD
|
|
170
|
+
effective_from: "2025-04-01"
|
|
171
|
+
source_url: https://prices.azure.com/api/retail/prices
|
|
172
|
+
verified: true
|
|
173
|
+
checked_on: "2026-11-13"
|
|
174
|
+
note: "Highest regional token meters (audio in 0.00792/1K, text 0.0033/1K, out 0.0132/1K); minute rate is a guard basis, not published. Retired 2026-10-15 per learn.microsoft.com/azure/foundry/openai/concepts/model-retirement-schedule, read 2026-11-18."
|
|
175
|
+
meters:
|
|
176
|
+
audio_input_per_minute: "0.00792"
|
|
177
|
+
audio_input_per_mtok: "7.92"
|
|
178
|
+
input_per_mtok: "3.3"
|
|
179
|
+
output_per_mtok: "13.2"
|
|
180
|
+
deprecation_date: "2026-10-15"
|
|
181
|
+
- id: azure_openai/gpt-4o-mini-transcribe
|
|
182
|
+
mode: transcription
|
|
183
|
+
currency: USD
|
|
184
|
+
effective_from: "2025-12-01"
|
|
185
|
+
source_url: https://prices.azure.com/api/retail/prices
|
|
186
|
+
verified: true
|
|
187
|
+
checked_on: "2026-11-13"
|
|
188
|
+
note: "Highest regional token meters (audio 3.96, text 1.65, out 6.60 per 1M); minute rate is a guard basis. Version 2025-12-15 retires 2027-06-15, 2025-03-20 retired 2026-10-15 (learn.microsoft.com/azure/foundry/openai/concepts/model-retirement-schedule, read 2026-11-18)."
|
|
189
|
+
meters:
|
|
190
|
+
audio_input_per_minute: "0.00396"
|
|
191
|
+
audio_input_per_mtok: "3.96"
|
|
192
|
+
input_per_mtok: "1.65"
|
|
193
|
+
output_per_mtok: "6.6"
|
|
194
|
+
deprecation_date: "2027-06-15"
|
|
195
|
+
|
|
196
|
+
# ===================== Azure Speech =====================
|
|
197
|
+
# Source: Azure Retail Prices API, serviceName 'Foundry Tools'/'MAI Models'. The azure.microsoft.com page shows blank prices.
|
|
198
|
+
- id: azure_speech/azure-speech
|
|
199
|
+
mode: transcription
|
|
200
|
+
currency: USD
|
|
201
|
+
effective_from: "2026-11-13"
|
|
202
|
+
source_url: https://prices.azure.com/api/retail/prices
|
|
203
|
+
verified: true
|
|
204
|
+
checked_on: "2026-11-13"
|
|
205
|
+
note: "Fast Transcription: highest regional meter 0.45 USD/h (min 0.36). Promo meter 0.10-0.125 ignored."
|
|
206
|
+
meters: { audio_input_per_minute: "0.0075" }
|
|
207
|
+
- id: azure_speech/mai-transcribe-2
|
|
208
|
+
mode: transcription
|
|
209
|
+
currency: USD
|
|
210
|
+
effective_from: "2026-10-01"
|
|
211
|
+
source_url: https://prices.azure.com/api/retail/prices
|
|
212
|
+
verified: true
|
|
213
|
+
checked_on: "2026-11-13"
|
|
214
|
+
note: "0.54 USD/h in every region (was launch price 0.10 in the earlier in-house guard; promo valid till 2026-12-31 per Azure page)."
|
|
215
|
+
meters: { audio_input_per_minute: "0.009" }
|
|
216
|
+
|
|
217
|
+
# ===================== Anthropic (direct API) =====================
|
|
218
|
+
# Page: platform.claude.com/docs/en/about-claude/pricing; limits from /docs/en/models/overview (max output 128K, context 1M;
|
|
219
|
+
# retirement not sooner than 2027-11-09 for Fable 5.2 and 2027-10-07 for Haiku 5.5, so no deprecation_date).
|
|
220
|
+
# cache_write_per_mtok holds the 1h cache-write price (dearer than the 5m price) because the schema has one meter.
|
|
221
|
+
- id: anthropic/claude-fable-5-2
|
|
222
|
+
mode: messages
|
|
223
|
+
currency: USD
|
|
224
|
+
effective_from: "2026-11-13"
|
|
225
|
+
source_url: https://platform.claude.com/docs/en/about-claude/pricing
|
|
226
|
+
verified: true
|
|
227
|
+
checked_on: "2026-11-13"
|
|
228
|
+
note: "Flagship. Input 8, cache write 5m 10 (1h 16 stored), cache hit 0.20 (0.025x), output 40."
|
|
229
|
+
meters: { input_per_mtok: "8", cached_input_per_mtok: "0.2", cache_write_per_mtok: "16", output_per_mtok: "40" }
|
|
230
|
+
limits: { max_output_tokens: 128000, context_window: 1000000 }
|
|
231
|
+
- id: anthropic/claude-haiku-5-5
|
|
232
|
+
mode: messages
|
|
233
|
+
currency: USD
|
|
234
|
+
effective_from: "2026-11-13"
|
|
235
|
+
source_url: https://platform.claude.com/docs/en/about-claude/pricing
|
|
236
|
+
verified: true
|
|
237
|
+
checked_on: "2026-11-13"
|
|
238
|
+
note: "Small. Prompts over 100,000 tokens: 5x on every meter (0.50/2.50 etc.). Cache write: 1h price 0.20 stored (5m 0.125)."
|
|
239
|
+
meters: { input_per_mtok: "0.1", cached_input_per_mtok: "0.01", cache_write_per_mtok: "0.2", output_per_mtok: "0.5" }
|
|
240
|
+
long_context: { threshold_input_tokens: 100000, input_multiplier: "5", output_multiplier: "5" }
|
|
241
|
+
limits: { max_output_tokens: 128000, context_window: 1000000 }
|
|
242
|
+
|
|
243
|
+
# ===================== AWS Bedrock (Converse), region us-east-1 =====================
|
|
244
|
+
# Source: AWS Price List bulk API (pricing.us-east-1.amazonaws.com/offers/v1.0/aws/AmazonBedrockFoundationModels/current),
|
|
245
|
+
# read 2026-11-13; the aws.amazon.com/bedrock/pricing page itself did not render the Anthropic tables to the fetcher.
|
|
246
|
+
# Model ids from platform.claude.com models overview. Cache write = 1h price (dearer than 5m).
|
|
247
|
+
- id: bedrock/global.anthropic.claude-fable-5-2
|
|
248
|
+
mode: messages
|
|
249
|
+
currency: USD
|
|
250
|
+
effective_from: "2026-11-01"
|
|
251
|
+
source_url: https://aws.amazon.com/bedrock/pricing/
|
|
252
|
+
verified: true
|
|
253
|
+
checked_on: "2026-11-13"
|
|
254
|
+
region: us-east-1
|
|
255
|
+
deployment_type: global
|
|
256
|
+
note: "Flagship, global inference profile price. Regional (non-global) us-east-1 would be +10% (8.8/44)."
|
|
257
|
+
meters: { input_per_mtok: "8", cached_input_per_mtok: "0.2", cache_write_per_mtok: "16", output_per_mtok: "40" }
|
|
258
|
+
limits: { max_output_tokens: 128000, context_window: 1000000 }
|
|
259
|
+
- id: bedrock/anthropic.claude-haiku-5-5
|
|
260
|
+
mode: messages
|
|
261
|
+
currency: USD
|
|
262
|
+
effective_from: "2026-10-01"
|
|
263
|
+
source_url: https://aws.amazon.com/bedrock/pricing/
|
|
264
|
+
verified: true
|
|
265
|
+
checked_on: "2026-11-13"
|
|
266
|
+
region: us-east-1
|
|
267
|
+
deployment_type: regional
|
|
268
|
+
note: "Small. Regional us-east-1 price (+10% over global, the dearer choice). Over 100,000 input tokens: 5x all meters."
|
|
269
|
+
meters: { input_per_mtok: "0.11", cached_input_per_mtok: "0.011", cache_write_per_mtok: "0.22", output_per_mtok: "0.55" }
|
|
270
|
+
long_context: { threshold_input_tokens: 100000, input_multiplier: "5", output_multiplier: "5" }
|
|
271
|
+
limits: { max_output_tokens: 128000, context_window: 1000000 }
|
|
272
|
+
|
|
273
|
+
# ===================== Google (Gemini API, AI Studio) =====================
|
|
274
|
+
# Page: ai.google.dev/gemini-api/docs/pricing (redirects to aistudio.google.com/docs/pricing), Standard paid tier.
|
|
275
|
+
# Vertex AI pricing page could not be read by the fetcher, so Vertex-specific prices are not encoded.
|
|
276
|
+
# max_output_tokens 65536: the models page says "65K / 1M" for Gemini 4 (ambiguous) and is silent for 3.5 Flash-Lite; the
|
|
277
|
+
# lower figure is used (it is sent as the cap, so cost is bounded). Context window 1M per models page for Gemini 4.
|
|
278
|
+
- id: google/gemini-4-carbon
|
|
279
|
+
mode: chat
|
|
280
|
+
currency: USD
|
|
281
|
+
effective_from: "2026-11-13"
|
|
282
|
+
source_url: https://ai.google.dev/gemini-api/docs/pricing
|
|
283
|
+
verified: true
|
|
284
|
+
checked_on: "2026-11-13"
|
|
285
|
+
note: "Flagship. Price valid through 2026-11-30; doubles from 2026-12-01 (see next entry). Implicit caching only."
|
|
286
|
+
meters: { input_per_mtok: "2", cached_input_per_mtok: "0.1", output_per_mtok: "10" }
|
|
287
|
+
limits: { max_output_tokens: 65536, context_window: 1000000 }
|
|
288
|
+
- id: google/gemini-4-carbon
|
|
289
|
+
mode: chat
|
|
290
|
+
currency: USD
|
|
291
|
+
effective_from: "2026-12-01"
|
|
292
|
+
source_url: https://ai.google.dev/gemini-api/docs/pricing
|
|
293
|
+
verified: true
|
|
294
|
+
checked_on: "2026-11-13"
|
|
295
|
+
note: "Announced price from 2026-12-01 ($4 input, $0.20 cached, $20 output)."
|
|
296
|
+
meters: { input_per_mtok: "4", cached_input_per_mtok: "0.2", output_per_mtok: "20" }
|
|
297
|
+
limits: { max_output_tokens: 65536, context_window: 1000000 }
|
|
298
|
+
- id: google/gemini-3.5-flash-lite
|
|
299
|
+
mode: chat
|
|
300
|
+
currency: USD
|
|
301
|
+
effective_from: "2026-11-13"
|
|
302
|
+
source_url: https://ai.google.dev/gemini-api/docs/pricing
|
|
303
|
+
verified: true
|
|
304
|
+
checked_on: "2026-11-13"
|
|
305
|
+
note: "Small/cheap current model (stable). Output includes thinking. Cache storage ($1/MTok/h) not encoded. Output limit not stated by Google."
|
|
306
|
+
meters: { input_per_mtok: "0.3", cached_input_per_mtok: "0.03", output_per_mtok: "2.5" }
|
|
307
|
+
limits: { max_output_tokens: 65536 }
|
|
308
|
+
|
|
309
|
+
# ===================== ElevenLabs (TTS, per 1,000 characters) =====================
|
|
310
|
+
- id: elevenlabs/eleven_v4
|
|
311
|
+
mode: speech
|
|
312
|
+
currency: USD
|
|
313
|
+
effective_from: "2026-11-10"
|
|
314
|
+
source_url: https://elevenlabs.io/pricing/api
|
|
315
|
+
verified: true
|
|
316
|
+
checked_on: "2026-11-13"
|
|
317
|
+
meters: { per_1k_characters: "0.08" }
|
|
318
|
+
- id: elevenlabs/eleven_v4_turbo
|
|
319
|
+
mode: speech
|
|
320
|
+
currency: USD
|
|
321
|
+
effective_from: "2026-11-10"
|
|
322
|
+
source_url: https://elevenlabs.io/pricing/api
|
|
323
|
+
verified: true
|
|
324
|
+
checked_on: "2026-11-13"
|
|
325
|
+
meters: { per_1k_characters: "0.04" }
|
|
326
|
+
- id: elevenlabs/eleven_multilingual_v2
|
|
327
|
+
mode: speech
|
|
328
|
+
currency: USD
|
|
329
|
+
effective_from: "2026-11-10"
|
|
330
|
+
source_url: https://elevenlabs.io/pricing/api
|
|
331
|
+
verified: true
|
|
332
|
+
checked_on: "2026-11-13"
|
|
333
|
+
meters: { per_1k_characters: "0.08" }
|
|
334
|
+
- id: elevenlabs/eleven_flash_v2_5
|
|
335
|
+
mode: speech
|
|
336
|
+
currency: USD
|
|
337
|
+
effective_from: "2026-11-10"
|
|
338
|
+
source_url: https://elevenlabs.io/pricing/api
|
|
339
|
+
verified: true
|
|
340
|
+
checked_on: "2026-11-13"
|
|
341
|
+
note: "Page lists 'Flash / Turbo' at 0.04 without naming v2.5."
|
|
342
|
+
meters: { per_1k_characters: "0.04" }
|
|
@@ -0,0 +1,135 @@
|
|
|
1
|
+
{
|
|
2
|
+
"$schema": "https://json-schema.org/draft/2020-12/schema",
|
|
3
|
+
"$id": "https://spendgate.invalid/schema/catalog-v1.json",
|
|
4
|
+
"title": "spendgate price catalog",
|
|
5
|
+
"description": "Prices and model metadata. All prices are decimal STRINGS (never floats). Unit is in the field name.",
|
|
6
|
+
"type": "object",
|
|
7
|
+
"additionalProperties": false,
|
|
8
|
+
"required": ["catalog_version", "models"],
|
|
9
|
+
"properties": {
|
|
10
|
+
"catalog_version": { "const": 1 },
|
|
11
|
+
"updated": { "$ref": "#/$defs/date" },
|
|
12
|
+
"providers": {
|
|
13
|
+
"type": "object",
|
|
14
|
+
"description": "Optional per-provider settlement rules, keyed by the provider part of a model id.",
|
|
15
|
+
"propertyNames": { "$ref": "#/$defs/slug" },
|
|
16
|
+
"additionalProperties": { "$ref": "#/$defs/settlement" }
|
|
17
|
+
},
|
|
18
|
+
"models": {
|
|
19
|
+
"type": "array",
|
|
20
|
+
"minItems": 1,
|
|
21
|
+
"items": { "$ref": "#/$defs/entry" }
|
|
22
|
+
}
|
|
23
|
+
},
|
|
24
|
+
"$defs": {
|
|
25
|
+
"date": { "type": "string", "format": "date", "pattern": "^[0-9]{4}-[0-9]{2}-[0-9]{2}$" },
|
|
26
|
+
"slug": { "type": "string", "pattern": "^[a-z][a-z0-9_]*$", "description": "Provider slug = adapter module name (openai, azure_openai, anthropic, bedrock, google, azure_speech, elevenlabs). Vertex is an auth mode of google, not a slug." },
|
|
27
|
+
"decimal": {
|
|
28
|
+
"type": "string",
|
|
29
|
+
"pattern": "^(0|[1-9][0-9]*)(\\.[0-9]+)?$",
|
|
30
|
+
"description": "Non-negative decimal as a string, e.g. \"0.012\"."
|
|
31
|
+
},
|
|
32
|
+
"multiplier": {
|
|
33
|
+
"type": "string",
|
|
34
|
+
"pattern": "^(0\\.[0-9]*[1-9][0-9]*|[1-9][0-9]*(\\.[0-9]+)?)$",
|
|
35
|
+
"description": "Strictly positive decimal string."
|
|
36
|
+
},
|
|
37
|
+
"settlement": {
|
|
38
|
+
"type": "object",
|
|
39
|
+
"additionalProperties": false,
|
|
40
|
+
"description": "Catalog data can never fail open: there is no zero-charge option. Fail-open knobs live only in spendgate.yaml under fail_open.",
|
|
41
|
+
"properties": {
|
|
42
|
+
"free_statuses": {
|
|
43
|
+
"type": "array",
|
|
44
|
+
"items": { "enum": [429, 529] },
|
|
45
|
+
"uniqueItems": true,
|
|
46
|
+
"default": [429],
|
|
47
|
+
"description": "Upstream HTTP statuses (throttling only) that mean the provider rejected the call before processing it: the attempt costs 0 and the request may fail over (core predicate not_processed). Default [429] for every provider; anthropic sets [429, 529]."
|
|
48
|
+
}
|
|
49
|
+
}
|
|
50
|
+
},
|
|
51
|
+
"long_context": {
|
|
52
|
+
"type": "object",
|
|
53
|
+
"additionalProperties": false,
|
|
54
|
+
"required": ["threshold_input_tokens", "input_multiplier", "output_multiplier"],
|
|
55
|
+
"description": "When a single request's input_tokens EXCEED the threshold, the whole request is priced at the multiplied rates.",
|
|
56
|
+
"properties": {
|
|
57
|
+
"threshold_input_tokens": { "type": "integer", "minimum": 1 },
|
|
58
|
+
"input_multiplier": { "$ref": "#/$defs/multiplier", "description": "Applies to input, cached input and cache-write meters (text and audio)." },
|
|
59
|
+
"output_multiplier": { "$ref": "#/$defs/multiplier", "description": "Applies to output and reasoning meters." }
|
|
60
|
+
}
|
|
61
|
+
},
|
|
62
|
+
"meters": {
|
|
63
|
+
"type": "object",
|
|
64
|
+
"additionalProperties": false,
|
|
65
|
+
"minProperties": 1,
|
|
66
|
+
"properties": {
|
|
67
|
+
"input_per_mtok": { "$ref": "#/$defs/decimal" },
|
|
68
|
+
"output_per_mtok": { "$ref": "#/$defs/decimal" },
|
|
69
|
+
"cached_input_per_mtok": { "$ref": "#/$defs/decimal" },
|
|
70
|
+
"cache_write_per_mtok": { "$ref": "#/$defs/decimal" },
|
|
71
|
+
"reasoning_per_mtok": { "$ref": "#/$defs/decimal" },
|
|
72
|
+
"audio_input_per_mtok": { "$ref": "#/$defs/decimal" },
|
|
73
|
+
"cached_audio_input_per_mtok": { "$ref": "#/$defs/decimal" },
|
|
74
|
+
"audio_output_per_mtok": { "$ref": "#/$defs/decimal" },
|
|
75
|
+
"audio_input_per_minute": { "$ref": "#/$defs/decimal" },
|
|
76
|
+
"audio_granularity_seconds": { "type": "integer", "minimum": 1, "default": 1, "description": "Audio duration is rounded UP to a multiple of this many seconds (min one unit)." },
|
|
77
|
+
"per_1k_characters": { "$ref": "#/$defs/decimal" },
|
|
78
|
+
"per_image": { "$ref": "#/$defs/decimal" }
|
|
79
|
+
}
|
|
80
|
+
},
|
|
81
|
+
"entry": {
|
|
82
|
+
"type": "object",
|
|
83
|
+
"additionalProperties": false,
|
|
84
|
+
"required": ["id", "mode", "currency", "effective_from", "source_url", "verified", "meters"],
|
|
85
|
+
"properties": {
|
|
86
|
+
"id": { "type": "string", "pattern": "^[a-z][a-z0-9_]*/[A-Za-z0-9][A-Za-z0-9._:/-]*$", "description": "provider-slug/model, as the caller names it." },
|
|
87
|
+
"mode": { "enum": ["chat", "responses", "messages", "embedding", "transcription", "speech"] },
|
|
88
|
+
"currency": { "enum": ["USD", "EUR"] },
|
|
89
|
+
"effective_from": { "$ref": "#/$defs/date" },
|
|
90
|
+
"source_url": { "type": "string", "format": "uri", "pattern": "^https://" },
|
|
91
|
+
"verified": { "type": "boolean", "description": "true only if a person compared every price to source_url on checked_on." },
|
|
92
|
+
"checked_on": { "$ref": "#/$defs/date" },
|
|
93
|
+
"note": { "type": "string", "maxLength": 300 },
|
|
94
|
+
"deployment_type": { "enum": ["global", "data_zone", "regional", "standard", "batch", "flex", "priority", "provisioned"] },
|
|
95
|
+
"region": { "type": "string", "pattern": "^[a-z0-9-]+$" },
|
|
96
|
+
"price_multiplier": { "$ref": "#/$defs/multiplier", "default": "1", "description": "Applied to every meter after long-context, before FX." },
|
|
97
|
+
"meters": { "$ref": "#/$defs/meters" },
|
|
98
|
+
"long_context": { "$ref": "#/$defs/long_context" },
|
|
99
|
+
"limits": {
|
|
100
|
+
"type": "object",
|
|
101
|
+
"additionalProperties": false,
|
|
102
|
+
"properties": {
|
|
103
|
+
"max_output_tokens": { "type": "integer", "minimum": 1 },
|
|
104
|
+
"context_window": { "type": "integer", "minimum": 1 }
|
|
105
|
+
}
|
|
106
|
+
},
|
|
107
|
+
"deprecation_date": { "$ref": "#/$defs/date", "description": "First day the model is refused (the provider's shutdown date). Unset means no known end." }
|
|
108
|
+
},
|
|
109
|
+
"allOf": [
|
|
110
|
+
{
|
|
111
|
+
"if": { "properties": { "mode": { "enum": ["chat", "responses", "messages"] } } },
|
|
112
|
+
"then": {
|
|
113
|
+
"required": ["limits"],
|
|
114
|
+
"properties": {
|
|
115
|
+
"meters": { "required": ["input_per_mtok", "output_per_mtok"] },
|
|
116
|
+
"limits": { "required": ["max_output_tokens"] }
|
|
117
|
+
}
|
|
118
|
+
}
|
|
119
|
+
},
|
|
120
|
+
{
|
|
121
|
+
"if": { "properties": { "mode": { "const": "embedding" } } },
|
|
122
|
+
"then": { "properties": { "meters": { "required": ["input_per_mtok"] } } }
|
|
123
|
+
},
|
|
124
|
+
{
|
|
125
|
+
"if": { "properties": { "mode": { "const": "transcription" } } },
|
|
126
|
+
"then": { "properties": { "meters": { "anyOf": [{ "required": ["audio_input_per_minute"] }, { "required": ["audio_input_per_mtok"] }] } } }
|
|
127
|
+
},
|
|
128
|
+
{
|
|
129
|
+
"if": { "properties": { "mode": { "const": "speech" } } },
|
|
130
|
+
"then": { "properties": { "meters": { "anyOf": [{ "required": ["per_1k_characters"] }, { "required": ["audio_output_per_mtok"] }] } } }
|
|
131
|
+
}
|
|
132
|
+
]
|
|
133
|
+
}
|
|
134
|
+
}
|
|
135
|
+
}
|
spendgate/cli/client.py
ADDED
|
@@ -0,0 +1,58 @@
|
|
|
1
|
+
"""The CLI's one HTTP client: a real admin API over TCP, or the admin app in-process (`--local`)."""
|
|
2
|
+
|
|
3
|
+
import httpx
|
|
4
|
+
|
|
5
|
+
type Json = dict[str, object]
|
|
6
|
+
|
|
7
|
+
EXIT_OK = 0
|
|
8
|
+
EXIT_REFUSED = 1 # the server understood and said no: invalid_request, not_found, conflict
|
|
9
|
+
EXIT_USAGE = 2 # argparse: bad command line
|
|
10
|
+
EXIT_CONFIG = 3 # the configuration file is invalid or missing
|
|
11
|
+
EXIT_AUTH = 4 # no admin key, or unauthorized / forbidden
|
|
12
|
+
EXIT_UNAVAILABLE = 5 # cannot reach the API, ledger unavailable, or a server error
|
|
13
|
+
_AUTH_STATUS = (401, 403)
|
|
14
|
+
_SERVER_ERROR = 500
|
|
15
|
+
|
|
16
|
+
|
|
17
|
+
class CliError(Exception):
|
|
18
|
+
"""Prints `message` on stderr and exits with `code`."""
|
|
19
|
+
|
|
20
|
+
def __init__(self, code: int, message: str) -> None:
|
|
21
|
+
super().__init__(message)
|
|
22
|
+
self.code = code
|
|
23
|
+
|
|
24
|
+
|
|
25
|
+
def _exit_code(status: int) -> int:
|
|
26
|
+
if status in _AUTH_STATUS:
|
|
27
|
+
return EXIT_AUTH
|
|
28
|
+
return EXIT_UNAVAILABLE if status >= _SERVER_ERROR else EXIT_REFUSED
|
|
29
|
+
|
|
30
|
+
|
|
31
|
+
class Client:
|
|
32
|
+
def __init__(self, http: httpx.AsyncClient) -> None:
|
|
33
|
+
self._http = http
|
|
34
|
+
|
|
35
|
+
async def call(
|
|
36
|
+
self,
|
|
37
|
+
method: str,
|
|
38
|
+
path: str,
|
|
39
|
+
*,
|
|
40
|
+
query: dict[str, str] | None = None,
|
|
41
|
+
body: Json | None = None,
|
|
42
|
+
) -> Json:
|
|
43
|
+
"""The decoded JSON answer (empty for 204); any failure raises `CliError`."""
|
|
44
|
+
try:
|
|
45
|
+
r = await self._http.request(method, path, params=query, json=body)
|
|
46
|
+
except httpx.TransportError:
|
|
47
|
+
raise CliError(EXIT_UNAVAILABLE, "cannot reach the admin API") from None
|
|
48
|
+
if r.status_code < 400: # noqa: PLR2004
|
|
49
|
+
decoded: Json = r.json() if r.content else {}
|
|
50
|
+
return decoded
|
|
51
|
+
try:
|
|
52
|
+
err = r.json()["error"]
|
|
53
|
+
text = f"{err['code']}: {err['message']}" + (
|
|
54
|
+
f" (param {err['param']})" if err.get("param") else ""
|
|
55
|
+
)
|
|
56
|
+
except (ValueError, KeyError, TypeError):
|
|
57
|
+
text = f"HTTP {r.status_code}"
|
|
58
|
+
raise CliError(_exit_code(r.status_code), text)
|