@ssweens/pi-vertex 1.1.13 → 1.4.0

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.
package/models/claude.ts CHANGED
@@ -11,7 +11,192 @@
11
11
  import type { VertexModelConfig } from "../types.js";
12
12
 
13
13
  export const CLAUDE_MODELS: VertexModelConfig[] = [
14
- // Claude Fable 5 (Mythos-class)
14
+ // Claude 5.5 series
15
+ // Note: Claude 5 / 5.5 / Fable models on Vertex require provider data sharing
16
+ // to be enabled on the publisher model config, otherwise requests fail with HTTP 403.
17
+ // Fable 5.1 and 5.2 also require the Advanced AI Safety Addendum in Model Garden.
18
+ {
19
+ id: "claude-opus-5-5",
20
+ name: "Claude Opus 5.5",
21
+ apiId: "claude-opus-5-5",
22
+ publisher: "anthropic",
23
+ endpointType: "maas",
24
+ contextWindow: 1000000,
25
+ maxTokens: 128000,
26
+ input: ["text", "image"],
27
+ reasoning: true,
28
+ tools: true,
29
+ cost: {
30
+ input: 4.00,
31
+ output: 20.00,
32
+ cacheRead: 0.20,
33
+ cacheWrite: 5.00,
34
+ },
35
+ costRegional: {
36
+ input: 4.40,
37
+ output: 22.00,
38
+ cacheRead: 0.22,
39
+ cacheWrite: 5.50,
40
+ },
41
+ region: "global",
42
+ },
43
+ {
44
+ id: "claude-sonnet-5-5",
45
+ name: "Claude Sonnet 5.5",
46
+ apiId: "claude-sonnet-5-5",
47
+ publisher: "anthropic",
48
+ endpointType: "maas",
49
+ contextWindow: 1000000,
50
+ maxTokens: 128000,
51
+ input: ["text", "image"],
52
+ reasoning: true,
53
+ tools: true,
54
+ cost: {
55
+ input: 2.00,
56
+ output: 10.00,
57
+ cacheRead: 0.20,
58
+ cacheWrite: 2.50,
59
+ },
60
+ costRegional: {
61
+ input: 2.20,
62
+ output: 11.00,
63
+ cacheRead: 0.22,
64
+ cacheWrite: 2.75,
65
+ },
66
+ region: "global",
67
+ },
68
+ // GA 2026-10-07. Prices are the <=100K-input tier: at >=100K input tokens
69
+ // every token bills at 5x ($0.50/$2.50 global), which this table does not model.
70
+ {
71
+ id: "claude-haiku-5-5",
72
+ name: "Claude Haiku 5.5",
73
+ apiId: "claude-haiku-5-5",
74
+ publisher: "anthropic",
75
+ endpointType: "maas",
76
+ contextWindow: 1000000,
77
+ maxTokens: 128000,
78
+ input: ["text", "image"],
79
+ reasoning: true,
80
+ tools: true,
81
+ cost: {
82
+ input: 0.10,
83
+ output: 0.50,
84
+ cacheRead: 0.01,
85
+ cacheWrite: 0.125,
86
+ },
87
+ costRegional: {
88
+ input: 0.11,
89
+ output: 0.55,
90
+ cacheRead: 0.011,
91
+ cacheWrite: 0.1375,
92
+ },
93
+ region: "global",
94
+ },
95
+
96
+ // Claude 5 series
97
+ {
98
+ id: "claude-opus-5",
99
+ name: "Claude Opus 5",
100
+ apiId: "claude-opus-5",
101
+ publisher: "anthropic",
102
+ endpointType: "maas",
103
+ contextWindow: 1000000,
104
+ maxTokens: 128000,
105
+ input: ["text", "image"],
106
+ reasoning: true,
107
+ tools: true,
108
+ cost: {
109
+ input: 5.00,
110
+ output: 25.00,
111
+ cacheRead: 0.50,
112
+ cacheWrite: 6.25,
113
+ },
114
+ costRegional: {
115
+ input: 5.50,
116
+ output: 27.50,
117
+ cacheRead: 0.55,
118
+ cacheWrite: 6.875,
119
+ },
120
+ region: "global",
121
+ },
122
+ {
123
+ id: "claude-sonnet-5",
124
+ name: "Claude Sonnet 5",
125
+ apiId: "claude-sonnet-5",
126
+ publisher: "anthropic",
127
+ endpointType: "maas",
128
+ contextWindow: 1000000,
129
+ maxTokens: 128000,
130
+ input: ["text", "image"],
131
+ reasoning: true,
132
+ tools: true,
133
+ cost: {
134
+ input: 2.00,
135
+ output: 10.00,
136
+ cacheRead: 0.20,
137
+ cacheWrite: 2.50,
138
+ },
139
+ costRegional: {
140
+ input: 2.20,
141
+ output: 11.00,
142
+ cacheRead: 0.22,
143
+ cacheWrite: 2.75,
144
+ },
145
+ region: "global",
146
+ },
147
+
148
+ // Claude Fable (Mythos-class)
149
+ // Fable 5.2: GA 2026-11-09. Fast mode (`speed: "fast"`, 2x price) is not used here.
150
+ {
151
+ id: "claude-fable-5-2",
152
+ name: "Claude Fable 5.2",
153
+ apiId: "claude-fable-5-2",
154
+ publisher: "anthropic",
155
+ endpointType: "maas",
156
+ contextWindow: 1000000,
157
+ maxTokens: 128000,
158
+ input: ["text", "image"],
159
+ reasoning: true,
160
+ tools: true,
161
+ cost: {
162
+ input: 8.00,
163
+ output: 40.00,
164
+ cacheRead: 0.20,
165
+ cacheWrite: 10.00,
166
+ },
167
+ costRegional: {
168
+ input: 8.80,
169
+ output: 44.00,
170
+ cacheRead: 0.22,
171
+ cacheWrite: 11.00,
172
+ },
173
+ region: "global",
174
+ },
175
+ {
176
+ id: "claude-fable-5-1",
177
+ name: "Claude Fable 5.1",
178
+ apiId: "claude-fable-5-1",
179
+ publisher: "anthropic",
180
+ endpointType: "maas",
181
+ contextWindow: 1000000,
182
+ maxTokens: 128000,
183
+ input: ["text", "image"],
184
+ reasoning: true,
185
+ tools: true,
186
+ cost: {
187
+ input: 10.00,
188
+ output: 50.00,
189
+ cacheRead: 0.25,
190
+ cacheWrite: 12.50,
191
+ },
192
+ costRegional: {
193
+ input: 11.00,
194
+ output: 55.00,
195
+ cacheRead: 0.275,
196
+ cacheWrite: 13.75,
197
+ },
198
+ region: "global",
199
+ },
15
200
  {
16
201
  id: "claude-fable-5",
17
202
  name: "Claude Fable 5",
@@ -20,7 +205,7 @@ export const CLAUDE_MODELS: VertexModelConfig[] = [
20
205
  endpointType: "maas",
21
206
  contextWindow: 1000000,
22
207
  maxTokens: 128000,
23
- input: ["text", "image", "file"],
208
+ input: ["text", "image"],
24
209
  reasoning: true,
25
210
  tools: true,
26
211
  cost: {
@@ -282,24 +467,5 @@ export const CLAUDE_MODELS: VertexModelConfig[] = [
282
467
  region: "global",
283
468
  },
284
469
 
285
- // Claude 3.5 series
286
- {
287
- id: "claude-3-5-sonnet-v2",
288
- name: "Claude 3.5 Sonnet v2",
289
- apiId: "claude-3-5-sonnet-v2@20241022",
290
- publisher: "anthropic",
291
- endpointType: "maas",
292
- contextWindow: 200000,
293
- maxTokens: 8192,
294
- input: ["text", "image"],
295
- reasoning: false,
296
- tools: true,
297
- cost: {
298
- input: 3.00,
299
- output: 15.00,
300
- cacheRead: 0.30,
301
- cacheWrite: 3.75,
302
- },
303
- region: "global",
304
- },
470
+ // Claude 3.5 Sonnet v2 was shut down on Vertex on 2026-02-19.
305
471
  ];
package/models/gemini.ts CHANGED
@@ -3,11 +3,186 @@
3
3
  * Source: https://docs.cloud.google.com/vertex-ai/generative-ai/docs/models
4
4
  * Pricing: https://cloud.google.com/vertex-ai/generative-ai/pricing
5
5
  * All prices per 1M tokens (standard tier, <= 200K input tokens)
6
+ * Verified against the Vertex AI model catalog on 2026-10-13; Gemini 4 entries
7
+ * re-verified against the Gemini Enterprise Agent Platform catalog on 2026-11-20.
6
8
  */
7
9
 
10
+ import type { ThinkingLevelMap } from "@earendil-works/pi-ai";
8
11
  import type { VertexModelConfig } from "../types.js";
9
12
 
13
+ // Gemini 4 rejects MINIMAL (HTTP 400): hide it in Pi; streaming/gemini.ts also floors it to LOW.
14
+ const GEMINI_4_THINKING: ThinkingLevelMap = { minimal: null };
15
+
10
16
  export const GEMINI_MODELS: VertexModelConfig[] = [
17
+ // --- Gemini 4 ---
18
+ // Unified 1M input+output limit. "Long decode" extends output to 1M tokens;
19
+ // 65536 is the per-segment streaming limit, consistent with the rest of the catalog.
20
+ // Implicit caching only; durable caching is on by default.
21
+
22
+ // GA 2026-11-18, Google Frontier Model. LOW/MEDIUM/HIGH (default)/XHIGH.
23
+ // Single-segment output cap on GenerateContent v1; input and output share 1M.
24
+ {
25
+ id: "gemini-4-neon",
26
+ name: "Gemini 4 Neon",
27
+ apiId: "gemini-4-neon",
28
+ publisher: "google",
29
+ endpointType: "gemini",
30
+ contextWindow: 1048576,
31
+ maxTokens: 65536,
32
+ input: ["text", "image"],
33
+ reasoning: true,
34
+ tools: true,
35
+ cost: {
36
+ input: 3.00,
37
+ output: 15.00,
38
+ cacheRead: 0.15,
39
+ cacheWrite: 0,
40
+ },
41
+ costRegional: {
42
+ input: 3.30,
43
+ output: 16.50,
44
+ cacheRead: 0.165,
45
+ cacheWrite: 0,
46
+ },
47
+ region: "global",
48
+ thinkingLevelMap: { ...GEMINI_4_THINKING, xhigh: "xhigh" },
49
+ },
50
+
51
+ // GA 2026-11-12, short-term release (45-day retirement notice). Thinking levels
52
+ // LOW/MEDIUM/HIGH (default)/XHIGH.
53
+ // Cache writes ($1.875/1M) bill from 2027-01-12; Gemini streams report no
54
+ // cache-write tokens, so cacheWrite stays 0.
55
+ {
56
+ id: "gemini-4-hydrogen",
57
+ name: "Gemini 4 Hydrogen",
58
+ apiId: "gemini-4-hydrogen",
59
+ publisher: "google",
60
+ endpointType: "gemini",
61
+ contextWindow: 1048576,
62
+ maxTokens: 65536,
63
+ input: ["text", "image"],
64
+ reasoning: true,
65
+ tools: true,
66
+ cost: {
67
+ input: 1.50,
68
+ output: 7.50,
69
+ cacheRead: 0.075,
70
+ cacheWrite: 0,
71
+ },
72
+ region: "global",
73
+ thinkingLevelMap: { ...GEMINI_4_THINKING, xhigh: "xhigh" },
74
+ },
75
+
76
+ // GA 2026-11-10, Google Frontier Model (verified enterprise customers in select
77
+ // regions). Thinking levels LOW/MEDIUM (default)/HIGH. Introductory pricing
78
+ // through 2026-11-30; rises to $4.00/$20.00 ($0.20 cached) after.
79
+ {
80
+ id: "gemini-4-carbon",
81
+ name: "Gemini 4 Carbon",
82
+ apiId: "gemini-4-carbon",
83
+ publisher: "google",
84
+ endpointType: "gemini",
85
+ contextWindow: 1048576,
86
+ maxTokens: 65536,
87
+ input: ["text", "image"],
88
+ reasoning: true,
89
+ tools: true,
90
+ cost: {
91
+ input: 2.00,
92
+ output: 10.00,
93
+ cacheRead: 0.10,
94
+ cacheWrite: 0,
95
+ },
96
+ region: "global",
97
+ thinkingLevelMap: GEMINI_4_THINKING,
98
+ },
99
+
100
+ // GA 2026-10-13. Thinking levels LOW/MEDIUM (default)/HIGH. Introductory
101
+ // pricing through 2026-11-30; rises to $4.00/$20.00 ($0.20 cached) after.
102
+ {
103
+ id: "gemini-4-argon",
104
+ name: "Gemini 4 Argon",
105
+ apiId: "gemini-4-argon",
106
+ publisher: "google",
107
+ endpointType: "gemini",
108
+ contextWindow: 1048576,
109
+ maxTokens: 65536,
110
+ input: ["text", "image"],
111
+ reasoning: true,
112
+ tools: true,
113
+ cost: {
114
+ input: 2.00,
115
+ output: 10.00,
116
+ cacheRead: 0.10,
117
+ cacheWrite: 0,
118
+ },
119
+ region: "global",
120
+ thinkingLevelMap: GEMINI_4_THINKING,
121
+ },
122
+
123
+ // --- Gemini 3.8 (GA, 2026-09-02) ---
124
+ {
125
+ id: "gemini-3.8-flash",
126
+ name: "Gemini 3.8 Flash",
127
+ apiId: "gemini-3.8-flash",
128
+ publisher: "google",
129
+ endpointType: "gemini",
130
+ contextWindow: 1048576,
131
+ maxTokens: 65536,
132
+ input: ["text", "image"],
133
+ reasoning: true,
134
+ tools: true,
135
+ cost: {
136
+ input: 0.75,
137
+ output: 3.75,
138
+ cacheRead: 0.075,
139
+ cacheWrite: 0,
140
+ },
141
+ region: "global",
142
+ },
143
+
144
+ // --- Gemini 3.7 (GA, 2026-08-13) ---
145
+ {
146
+ id: "gemini-3.7-flash",
147
+ name: "Gemini 3.7 Flash",
148
+ apiId: "gemini-3.7-flash",
149
+ publisher: "google",
150
+ endpointType: "gemini",
151
+ contextWindow: 1048576,
152
+ maxTokens: 65536,
153
+ input: ["text", "image"],
154
+ reasoning: true,
155
+ tools: true,
156
+ cost: {
157
+ input: 0.75,
158
+ output: 3.75,
159
+ cacheRead: 0.075,
160
+ cacheWrite: 0,
161
+ },
162
+ region: "global",
163
+ },
164
+
165
+ // --- Gemini 3.6 (GA, 2026-07-21) ---
166
+ {
167
+ id: "gemini-3.6-flash",
168
+ name: "Gemini 3.6 Flash",
169
+ apiId: "gemini-3.6-flash",
170
+ publisher: "google",
171
+ endpointType: "gemini",
172
+ contextWindow: 1048576,
173
+ maxTokens: 65536,
174
+ input: ["text", "image"],
175
+ reasoning: true,
176
+ tools: true,
177
+ cost: {
178
+ input: 1.50,
179
+ output: 7.50,
180
+ cacheRead: 0.15,
181
+ cacheWrite: 0,
182
+ },
183
+ region: "global",
184
+ },
185
+
11
186
  // --- Gemini 3.5 (GA) ---
12
187
  {
13
188
  id: "gemini-3.5-flash",
@@ -28,8 +203,27 @@ export const GEMINI_MODELS: VertexModelConfig[] = [
28
203
  },
29
204
  region: "global",
30
205
  },
206
+ {
207
+ id: "gemini-3.5-flash-lite",
208
+ name: "Gemini 3.5 Flash Lite",
209
+ apiId: "gemini-3.5-flash-lite",
210
+ publisher: "google",
211
+ endpointType: "gemini",
212
+ contextWindow: 1048576,
213
+ maxTokens: 65536,
214
+ input: ["text", "image"],
215
+ reasoning: true,
216
+ tools: true,
217
+ cost: {
218
+ input: 0.30,
219
+ output: 2.50,
220
+ cacheRead: 0.03,
221
+ cacheWrite: 0,
222
+ },
223
+ region: "global",
224
+ },
31
225
 
32
- // --- Gemini 3.1 (Preview) ---
226
+ // --- Gemini 3.1 (Pro still Preview; Flash Lite now GA) ---
33
227
  {
34
228
  id: "gemini-3.1-pro",
35
229
  name: "Gemini 3.1 Pro",
@@ -52,11 +246,11 @@ export const GEMINI_MODELS: VertexModelConfig[] = [
52
246
  {
53
247
  id: "gemini-3.1-flash-lite",
54
248
  name: "Gemini 3.1 Flash Lite",
55
- apiId: "gemini-3.1-flash-lite-preview",
249
+ apiId: "gemini-3.1-flash-lite",
56
250
  publisher: "google",
57
251
  endpointType: "gemini",
58
252
  contextWindow: 1048576,
59
- maxTokens: 65535,
253
+ maxTokens: 65536,
60
254
  input: ["text", "image"],
61
255
  reasoning: true,
62
256
  tools: true,
@@ -149,43 +343,7 @@ export const GEMINI_MODELS: VertexModelConfig[] = [
149
343
  region: "global",
150
344
  },
151
345
 
152
- // --- Gemini 2.0 (GA) ---
153
- {
154
- id: "gemini-2.0-flash",
155
- name: "Gemini 2.0 Flash",
156
- apiId: "gemini-2.0-flash",
157
- publisher: "google",
158
- endpointType: "gemini",
159
- contextWindow: 1048576,
160
- maxTokens: 8192,
161
- input: ["text", "image"],
162
- reasoning: false,
163
- tools: true,
164
- cost: {
165
- input: 0.15,
166
- output: 0.60,
167
- cacheRead: 0,
168
- cacheWrite: 0,
169
- },
170
- region: "global",
171
- },
172
- {
173
- id: "gemini-2.0-flash-lite",
174
- name: "Gemini 2.0 Flash Lite",
175
- apiId: "gemini-2.0-flash-lite",
176
- publisher: "google",
177
- endpointType: "gemini",
178
- contextWindow: 1048576,
179
- maxTokens: 8192,
180
- input: ["text", "image"],
181
- reasoning: false,
182
- tools: true,
183
- cost: {
184
- input: 0.075,
185
- output: 0.30,
186
- cacheRead: 0,
187
- cacheWrite: 0,
188
- },
189
- region: "global",
190
- },
346
+ // Gemini 2.0 Flash / Flash Lite were shut down on Vertex AI on 2026-06-01.
347
+ // Migration path is gemini-2.5-flash / gemini-2.5-flash-lite, which are
348
+ // themselves retired on 2026-10-16.
191
349
  ];