File size: 18,885 Bytes
e93eb3d
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
32
33
34
35
36
37
38
39
40
41
42
43
44
45
46
47
48
49
50
51
52
53
54
55
56
57
58
59
60
61
62
63
64
65
66
67
68
69
70
71
72
73
74
75
76
77
78
79
80
81
82
83
84
85
86
87
88
89
90
91
92
93
94
95
96
97
98
99
100
101
102
103
104
105
106
107
108
109
110
111
112
113
114
115
116
117
118
119
120
121
122
123
124
125
126
127
128
129
130
131
132
133
134
135
136
137
138
139
140
141
142
143
144
145
146
147
148
149
150
151
152
153
154
155
156
157
158
159
160
161
162
163
164
165
166
167
168
169
170
171
172
173
174
175
176
177
178
179
180
181
182
183
184
185
186
187
188
189
190
191
192
193
194
195
196
197
198
199
200
201
202
203
204
205
206
207
208
209
210
211
212
213
214
215
216
217
218
219
220
221
222
223
224
225
226
227
228
229
230
231
232
233
234
235
236
237
238
239
240
241
242
243
244
245
246
247
248
249
250
251
252
253
254
255
256
257
258
259
260
261
262
263
264
265
266
267
268
269
270
271
272
273
274
275
276
277
278
279
280
281
282
283
284
285
286
287
288
289
290
291
292
293
294
295
296
297
298
299
300
301
302
303
304
305
306
307
308
309
310
311
312
313
314
315
316
317
318
319
320
321
322
323
324
325
326
327
328
329
330
331
332
333
334
335
336
337
338
339
340
341
342
343
344
345
346
347
348
349
350
351
352
353
354
355
356
357
358
359
360
361
362
363
364
365
366
367
368
369
370
371
372
373
374
375
376
377
378
379
380
381
382
383
384
385
386
387
388
389
390
391
392
393
394
395
396
397
398
399
400
401
402
403
404
405
406
407
408
409
410
411
412
413
414
415
416
417
418
419
420
421
422
423
424
425
426
427
428
429
430
431
432
433
434
435
436
437
438
439
440
441
442
443
444
445
446
447
448
449
450
451
452
453
454
455
456
457
458
459
460
461
462
463
464
465
466
467
468
469
470
471
472
473
474
475
476
477
478
479
480
481
482
483
484
import ast
import datetime
import os
from typing import List, Dict, Generator
from dotenv import load_dotenv

from tenacity import (
    retry,
    stop_after_attempt,
    wait_random_exponential,
)  # for exponential backoff

# Load environment variables from .env file
load_dotenv()


@retry(wait=wait_random_exponential(min=1, max=60), stop=stop_after_attempt(3))
def anthropic_completion_with_backoff(client, *args, **kwargs):
    return client.beta.prompt_caching.messages.create(*args, **kwargs)


def get_anthropic(model: str,
                  prompt: str,
                  temperature: float = 0,
                  max_tokens: int = 4096,
                  system: str = '',
                  chat_history: List[Dict] = None,
                  verbose=False) -> \
        Generator[dict, None, None]:
    model = model.replace('anthropic:', '')

    # https://docs.anthropic.com/en/docs/build-with-claude/prompt-caching
    import anthropic

    clawd_key = os.getenv('ANTHROPIC_API_KEY')
    clawd_client = anthropic.Anthropic(api_key=clawd_key) if clawd_key else None

    if chat_history is None:
        chat_history = []

    messages = []

    # Add conversation history, removing cache_control from all but the last two user messages
    for i, message in enumerate(chat_history):
        if message["role"] == "user":
            if i >= len(chat_history) - 3:  # Last two user messages
                messages.append(message)
            else:
                messages.append({
                    "role": "user",
                    "content": [{"type": "text", "text": message["content"][0]["text"]}]
                })
        else:
            messages.append(message)

    # Add the new user message
    messages.append({
        "role": "user",
        "content": [
            {
                "type": "text",
                "text": prompt,
                "cache_control": {"type": "ephemeral"}
            }
        ]
    })

    response = anthropic_completion_with_backoff(clawd_client,
                                                 model=model,
                                                 max_tokens=max_tokens,
                                                 temperature=temperature,
                                                 system=system,
                                                 messages=messages,
                                                 stream=True
                                                 )

    output_tokens = 0
    input_tokens = 0
    cache_creation_input_tokens = 0
    cache_read_input_tokens = 0
    for chunk in response:
        if chunk.type == "content_block_start":
            # This is where we might find usage info in the future
            pass
        elif chunk.type == "content_block_delta":
            yield dict(text=chunk.delta.text)
        elif chunk.type == "message_delta":
            output_tokens = dict(chunk.usage).get('output_tokens', 0)
        elif chunk.type == "message_start":
            usage = chunk.message.usage
            input_tokens = dict(usage).get('input_tokens', 0)
            cache_creation_input_tokens = dict(usage).get('cache_creation_input_tokens', 0)
            cache_read_input_tokens = dict(usage).get('cache_read_input_tokens', 0)
        else:
            if verbose:
                print("Unknown chunk type:", chunk.type)
                print("Chunk:", chunk)

    if verbose:
        # After streaming is complete, print the usage information
        print(f"Output tokens: {output_tokens}")
        print(f"Input tokens: {input_tokens}")
        print(f"Cache creation input tokens: {cache_creation_input_tokens}")
        print(f"Cache read input tokens: {cache_read_input_tokens}")
    yield dict(output_tokens=output_tokens, input_tokens=input_tokens,
               cache_creation_input_tokens=cache_creation_input_tokens,
               cache_read_input_tokens=cache_read_input_tokens)


@retry(wait=wait_random_exponential(min=1, max=60), stop=stop_after_attempt(3))
def openai_completion_with_backoff(client, *args, **kwargs):
    return client.chat.completions.create(*args, **kwargs)


def get_openai(model: str,
               prompt: str,
               temperature: float = 0,
               max_tokens: int = 4096,
               system: str = '',
               chat_history: List[Dict] = None,
               verbose=False) -> Generator[dict, None, None]:
    anthropic_models, openai_models, google_models, groq_models, azure_models, ollama = get_model_names()
    if model in ollama:
        model = model.replace('ollama:', '')
        openai_key = os.getenv('OLLAMA_OPENAI_API_KEY')
        openai_base_url = os.getenv('OLLAMA_OPENAI_BASE_URL', 'http://localhost:11434/v1/')
    else:
        model = model.replace('openai:', '')
        openai_key = os.getenv('OPENAI_API_KEY')
        openai_base_url = os.getenv('OPENAI_BASE_URL', 'https://api.openai.com/v1')

    from openai import OpenAI

    openai_client = OpenAI(api_key=openai_key, base_url=openai_base_url) if openai_key else None

    if chat_history is None:
        chat_history = []
    chat_history_copy = chat_history.copy()
    for mi, message in enumerate(chat_history_copy):
        if isinstance(message["content"], list):
            chat_history_copy[mi]["content"] = message["content"][0]["text"]
    chat_history = chat_history_copy

    messages = [{"role": "system", "content": system}] + chat_history + [{"role": "user", "content": prompt}]

    response = openai_completion_with_backoff(openai_client,
                                              model=model,
                                              messages=messages,
                                              temperature=temperature,
                                              max_tokens=max_tokens,
                                              stream=True,
                                              )

    output_tokens = 0
    input_tokens = 0
    for chunk in response:
        if chunk.choices[0].delta.content:
            yield dict(text=chunk.choices[0].delta.content)
        if chunk.usage:
            output_tokens = chunk.usage.completion_tokens
            input_tokens = chunk.usage.prompt_tokens

    if verbose:
        print(f"Output tokens: {output_tokens}")
        print(f"Input tokens: {input_tokens}")
    yield dict(output_tokens=output_tokens, input_tokens=input_tokens)


def openai_messages_to_gemini_history(messages):
    """Converts OpenAI messages to Gemini history format.

    Args:
        messages: A list of OpenAI messages, each with "role" and "content" keys.

    Returns:
        A list of dictionaries representing the chat history for Gemini.
    """
    history = []
    for message in messages:
        if isinstance(message["content"], list):
            message["content"] = message["content"][0]["text"]
        if message["role"] == "user":
            history.append({"role": "user", "parts": [{"text": message["content"]}]})
        elif message["role"] == "assistant":
            history.append({"role": "model", "parts": [{"text": message["content"]}]})
        # Optionally handle system messages if needed
        # elif message["role"] == "system":
        #     history.append({"role": "system", "parts": [{"text": message["content"]}]})

    return history


@retry(wait=wait_random_exponential(min=1, max=60), stop=stop_after_attempt(3))
def gemini_send_message_with_backoff(chat, prompt, stream=True):
    return chat.send_message(prompt, stream=stream)


@retry(wait=wait_random_exponential(min=1, max=60), stop=stop_after_attempt(3))
def gemini_generate_content_with_backoff(model, prompt, stream=True):
    return model.generate_content(prompt, stream=stream)


def get_google(model: str,
               prompt: str,
               temperature: float = 0,
               max_tokens: int = 4096,
               system: str = '',
               chat_history: List[Dict] = None,
               verbose=False) -> Generator[dict, None, None]:
    model = model.replace('google:', '').replace('gemini:', '')

    import google.generativeai as genai

    gemini_key = os.getenv("GEMINI_API_KEY")
    genai.configure(api_key=gemini_key)
    # Create the model
    generation_config = {
        "temperature": temperature,
        "top_p": 0.95,
        "top_k": 64,
        "max_output_tokens": max_tokens,
        "response_mime_type": "text/plain",
    }

    if chat_history is None:
        chat_history = []

    chat_history = chat_history.copy()
    chat_history = openai_messages_to_gemini_history(chat_history)

    # NOTE: assume want own control.  Too many false positives by Google.
    from google.generativeai.types import HarmCategory
    from google.generativeai.types import HarmBlockThreshold
    safety_settings = {
        HarmCategory.HARM_CATEGORY_DANGEROUS_CONTENT: HarmBlockThreshold.BLOCK_NONE,
        HarmCategory.HARM_CATEGORY_HATE_SPEECH: HarmBlockThreshold.BLOCK_NONE,
        HarmCategory.HARM_CATEGORY_HARASSMENT: HarmBlockThreshold.BLOCK_NONE,
        HarmCategory.HARM_CATEGORY_SEXUALLY_EXPLICIT: HarmBlockThreshold.BLOCK_NONE,
    }

    cache = None
    # disable cache for now until work into things well
    use_cache = False
    if use_cache and model == 'gemini-1.5-pro':
        from google.generativeai import caching
        # Estimate token count (this is a rough estimate, you may need a more accurate method)
        estimated_tokens = len(prompt.split()) + sum(len(msg['content'].split()) for msg in chat_history)

        if estimated_tokens > 32000:
            cache = caching.CachedContent.create(
                model=model,
                display_name=f'cache_{datetime.datetime.now().isoformat()}',
                system_instruction=system,
                contents=[prompt] + [msg['content'] for msg in chat_history],
                ttl=datetime.timedelta(minutes=5),  # Set an appropriate TTL.  Short for now for cost savings.
            )
            gemini_model = genai.GenerativeModel.from_cached_content(cached_content=cache)
        else:
            gemini_model = genai.GenerativeModel(model_name=model,
                                                 generation_config=generation_config,
                                                 safety_settings=safety_settings)
    else:
        gemini_model = genai.GenerativeModel(model_name=model,
                                             generation_config=generation_config,
                                             safety_settings=safety_settings)

    if cache:
        response = gemini_generate_content_with_backoff(gemini_model, prompt, stream=True)
    else:
        chat = gemini_model.start_chat(history=chat_history)
        response = gemini_send_message_with_backoff(chat, prompt, stream=True)

    output_tokens = 0
    input_tokens = 0
    cache_read_input_tokens = 0
    cache_creation_input_tokens = 0

    for chunk in response:
        if chunk.text:
            yield dict(text=chunk.text)
        if chunk.usage_metadata:
            output_tokens = chunk.usage_metadata.candidates_token_count
            input_tokens = chunk.usage_metadata.prompt_token_count
            cache_read_input_tokens = chunk.usage_metadata.cached_content_token_count
            cache_creation_input_tokens = 0  # This might need to be updated if available in the API

    if verbose:
        print(f"Output tokens: {output_tokens}")
        print(f"Input tokens: {input_tokens}")
        print(f"Cached tokens: {cache_read_input_tokens}")

    yield dict(output_tokens=output_tokens, input_tokens=input_tokens,
               cache_read_input_tokens=cache_read_input_tokens,
               cache_creation_input_tokens=cache_creation_input_tokens)


def delete_cache(cache):
    if cache:
        cache.delete()
        print(f"Cache {cache.display_name} deleted.")
    else:
        print("No cache to delete.")


def get_groq(model: str,
             prompt: str,
             temperature: float = 0,
             max_tokens: int = 4096,
             system: str = '',
             chat_history: List[Dict] = None,
             verbose=False) -> Generator[dict, None, None]:
    model = model.replace('groq:', '')

    from groq import Groq

    groq_key = os.getenv("GROQ_API_KEY")
    client = Groq(api_key=groq_key)

    if chat_history is None:
        chat_history = []

    chat_history = chat_history.copy()

    messages = [{"role": "system", "content": system}] + chat_history + [{"role": "user", "content": prompt}]

    stream = openai_completion_with_backoff(client,
                                            messages=messages,
                                            model=model,
                                            temperature=temperature,
                                            max_tokens=max_tokens,
                                            stream=True,
                                            )

    output_tokens = 0
    input_tokens = 0
    for chunk in stream:
        if chunk.choices[0].delta.content:
            yield dict(text=chunk.choices[0].delta.content)
        if chunk.usage:
            output_tokens = chunk.usage.completion_tokens
            input_tokens = chunk.usage.prompt_tokens

    if verbose:
        print(f"Output tokens: {output_tokens}")
        print(f"Input tokens: {input_tokens}")
    yield dict(output_tokens=output_tokens, input_tokens=input_tokens)


def get_openai_azure(model: str,
                     prompt: str,
                     temperature: float = 0,
                     max_tokens: int = 4096,
                     system: str = '',
                     chat_history: List[Dict] = None,
                     verbose=False) -> Generator[dict, None, None]:
    model = model.replace('azure:', '').replace('openai_azure:', '')

    from openai import AzureOpenAI

    azure_endpoint = os.getenv("AZURE_OPENAI_ENDPOINT")  # e.g. https://project.openai.azure.com
    azure_key = os.getenv("AZURE_OPENAI_API_KEY")
    azure_deployment = os.getenv("AZURE_OPENAI_DEPLOYMENT")  # i.e. deployment name with some models deployed
    azure_api_version = os.getenv('AZURE_OPENAI_API_VERSION', '2024-07-01-preview')
    assert azure_endpoint is not None, "Azure OpenAI endpoint not set"
    assert azure_key is not None, "Azure OpenAI API key not set"
    assert azure_deployment is not None, "Azure OpenAI deployment not set"

    client = AzureOpenAI(
        azure_endpoint=azure_endpoint,
        api_key=azure_key,
        api_version=azure_api_version,
        azure_deployment=azure_deployment,
    )

    if chat_history is None:
        chat_history = []

    messages = [{"role": "system", "content": system}] + chat_history + [{"role": "user", "content": prompt}]

    response = openai_completion_with_backoff(client,
                                              model=model,
                                              messages=messages,
                                              temperature=temperature,
                                              max_tokens=max_tokens,
                                              stream=True
                                              )

    output_tokens = 0
    input_tokens = 0
    for chunk in response:
        if chunk.choices and chunk.choices[0].delta.content:
            yield dict(text=chunk.choices[0].delta.content)
        if chunk.usage:
            output_tokens = chunk.usage.completion_tokens
            input_tokens = chunk.usage.prompt_tokens

    if verbose:
        print(f"Output tokens: {output_tokens}")
        print(f"Input tokens: {input_tokens}")
    yield dict(output_tokens=output_tokens, input_tokens=input_tokens)


def to_list(x):
    if x:
        try:
            ollama_model_list = ast.literal_eval(x)
            assert isinstance(ollama_model_list, list)
        except:
            x = [x]
    else:
        x = []
    return x


def get_model_names(secrets, on_hf_spaces=False):
    if not on_hf_spaces:
        secrets = os.environ
    if secrets.get('ANTHROPIC_API_KEY'):
        anthropic_models = ['claude-3-5-sonnet-20240620', 'claude-3-haiku-20240307', 'claude-3-opus-20240229']
    else:
        anthropic_models = []
    if secrets.get('OPENAI_API_KEY'):
        if os.getenv('OPENAI_MODEL_NAME'):
            openai_models = to_list(os.getenv('OPENAI_MODEL_NAME'))
        else:
            openai_models = ['gpt-4o', 'gpt-4-turbo-2024-04-09', 'gpt-4o-mini']
    else:
        openai_models = []
    if secrets.get('AZURE_OPENAI_API_KEY'):
        if os.getenv('AZURE_OPENAI_MODEL_NAME'):
            azure_models = to_list(os.getenv('AZURE_OPENAI_MODEL_NAME'))
        else:
            azure_models = ['gpt-4o', 'gpt-4-turbo-2024-04-09', 'gpt-4o-mini']
    else:
        azure_models = []
    if secrets.get('GEMINI_API_KEY'):
        google_models = ['gemini-1.5-pro-latest', 'gemini-1.5-flash-latest']
    else:
        google_models = []
    if secrets.get('GROQ_API_KEY'):
        groq_models = ['llama-3.1-70b-versatile',
                       'llama-3.1-8b-instant',
                       'llama3-groq-70b-8192-tool-use-preview',
                       'llama3-groq-8b-8192-tool-use-preview',
                       'mixtral-8x7b-32768']
    else:
        groq_models = []
    if secrets.get('OLLAMA_OPENAI_API_KEY'):
        ollama_model = os.environ['OLLAMA_OPENAI_MODEL_NAME']
        ollama_model = to_list(ollama_model)
    else:
        ollama_model = []

    groq_models = ['groq:' + x for x in groq_models]
    azure_models = ['azure:' + x for x in azure_models]
    openai_models = ['openai:' + x for x in openai_models]
    google_models = ['google:' + x for x in google_models]
    anthropic_models = ['anthropic:' + x for x in anthropic_models]
    ollama = ['ollama:' + x if 'ollama:' not in x else x for x in ollama_model]

    return anthropic_models, openai_models, google_models, groq_models, azure_models, ollama


def get_all_model_names(secrets, on_hf_spaces=False):
    anthropic_models, openai_models, google_models, groq_models, azure_models, ollama = get_model_names(secrets,
                                                                                                        on_hf_spaces=on_hf_spaces)
    return anthropic_models + openai_models + google_models + groq_models + azure_models + ollama


def get_model_api(model: str):
    if model.startswith('anthropic:'):
        return get_anthropic
    elif model.startswith('openai:') or model.startswith('ollama:'):
        return get_openai
    elif model.startswith('google:'):
        return get_google
    elif model.startswith('groq:'):
        return get_groq
    elif model.startswith('azure:'):
        return get_openai_azure
    else:
        raise ValueError(
            f"Unsupported model: {model}.  Ensure to add prefix (e.g. openai:, google:, groq:, azure:, ollama:, anthropic:)")