langfun 0.1.2.dev202501090804__py3-none-any.whl → 0.1.2.dev202501110804__py3-none-any.whl

langfun/core/llms/openai_test.py

@@ -13,102 +13,9 @@
 # limitations under the License.
 """Tests for OpenAI models."""
 
-from typing import Any
 import unittest
-from unittest import mock
-
 import langfun.core as lf
-from langfun.core import modalities as lf_modalities
 from langfun.core.llms import openai
-import pyglove as pg
-import requests
-
-
-def mock_chat_completion_request(url: str, json: dict[str, Any], **kwargs):
-  del url, kwargs
-  messages = json['messages']
-  if len(messages) > 1:
-    system_message = f' system={messages[0]["content"]}'
-  else:
-    system_message = ''
-
-  if 'response_format' in json:
-    response_format = f' format={json["response_format"]["type"]}'
-  else:
-    response_format = ''
-
-  choices = []
-  for k in range(json['n']):
-    if json.get('logprobs'):
-      logprobs = dict(
-          content=[
-              dict(
-                  token='chosen_token',
-                  logprob=0.5,
-                  top_logprobs=[
-                      dict(
-                          token=f'alternative_token_{i + 1}',
-                          logprob=0.1
-                      ) for i in range(3)
-                  ]
-              )
-          ]
-      )
-    else:
-      logprobs = None
-
-    choices.append(dict(
-        message=dict(
-            content=(
-                f'Sample {k} for message.{system_message}{response_format}'
-            )
-        ),
-        logprobs=logprobs,
-    ))
-  response = requests.Response()
-  response.status_code = 200
-  response._content = pg.to_json_str(
-      dict(
-          choices=choices,
-          usage=lf.LMSamplingUsage(
-              prompt_tokens=100,
-              completion_tokens=100,
-              total_tokens=200,
-          ),
-      )
-  ).encode()
-  return response
-
-
-def mock_chat_completion_request_vision(
-    url: str, json: dict[str, Any], **kwargs
-):
-  del url, kwargs
-  choices = []
-  urls = [
-      c['image_url']['url']
-      for c in json['messages'][0]['content'] if c['type'] == 'image_url'
-  ]
-  for k in range(json['n']):
-    choices.append(pg.Dict(
-        message=pg.Dict(
-            content=f'Sample {k} for message: {"".join(urls)}'
-        ),
-        logprobs=None,
-    ))
-  response = requests.Response()
-  response.status_code = 200
-  response._content = pg.to_json_str(
-      dict(
-          choices=choices,
-          usage=lf.LMSamplingUsage(
-              prompt_tokens=100,
-              completion_tokens=100,
-              total_tokens=200,
-          ),
-      )
-  ).encode()
-  return response
 
 
 class OpenAITest(unittest.TestCase):
@@ -130,6 +37,15 @@ class OpenAITest(unittest.TestCase):
         openai.Gpt35(api_key='test_key').resource_id, 'OpenAI(text-davinci-003)'
     )
 
+  def test_headers(self):
+    self.assertEqual(
+        openai.Gpt35(api_key='test_key').headers,
+        {
+            'Content-Type': 'application/json',
+            'Authorization': 'Bearer test_key',
+        },
+    )
+
   def test_max_concurrency(self):
     self.assertGreater(openai.Gpt35(api_key='test_key').max_concurrency, 0)
 
@@ -156,340 +72,14 @@ class OpenAITest(unittest.TestCase):
           )
       )
 
-  def test_call_chat_completion(self):
-    with mock.patch('requests.Session.post') as mock_request:
-      mock_request.side_effect = mock_chat_completion_request
-      lm = openai.OpenAI(
-          model='gpt-4',
-          api_key='test_key',
-          organization='my_org',
-          project='my_project'
-      )
-      self.assertEqual(
-          lm('hello', sampling_options=lf.LMSamplingOptions(n=2)),
-          'Sample 0 for message.',
-      )
-
-  def test_call_chat_completion_with_logprobs(self):
-    with mock.patch('requests.Session.post') as mock_request:
-      mock_request.side_effect = mock_chat_completion_request
-      lm = openai.OpenAI(
-          model='gpt-4',
-          api_key='test_key',
-          organization='my_org',
-          project='my_project'
-      )
-      results = lm.sample(['hello'], logprobs=True)
-      self.assertEqual(len(results), 1)
-      self.assertEqual(
-          results[0],
-          lf.LMSamplingResult(
-              [
-                  lf.LMSample(
-                      response=lf.AIMessage(
-                          text='Sample 0 for message.',
-                          metadata={
-                              'score': 0.0,
-                              'logprobs': [(
-                                  'chosen_token',
-                                  0.5,
-                                  [
-                                      ('alternative_token_1', 0.1),
-                                      ('alternative_token_2', 0.1),
-                                      ('alternative_token_3', 0.1),
-                                  ],
-                              )],
-                              'is_cached': False,
-                              'usage': lf.LMSamplingUsage(
-                                  prompt_tokens=100,
-                                  completion_tokens=100,
-                                  total_tokens=200,
-                                  estimated_cost=0.009,
-                              ),
-                          },
-                          tags=['lm-response'],
-                      ),
-                      logprobs=[(
-                          'chosen_token',
-                          0.5,
-                          [
-                              ('alternative_token_1', 0.1),
-                              ('alternative_token_2', 0.1),
-                              ('alternative_token_3', 0.1),
-                          ],
-                      )],
-                  )
-              ],
-              usage=lf.LMSamplingUsage(
-                  prompt_tokens=100,
-                  completion_tokens=100,
-                  total_tokens=200,
-                  estimated_cost=0.009,
-              ),
-          ),
-      )
-
-  def test_call_chat_completion_vision(self):
-    with mock.patch('requests.Session.post') as mock_request:
-      mock_request.side_effect = mock_chat_completion_request_vision
-      lm_1 = openai.Gpt4Turbo(api_key='test_key')
-      lm_2 = openai.Gpt4VisionPreview(api_key='test_key')
-      for lm in (lm_1, lm_2):
-        self.assertEqual(
-            lm(
-                lf.UserMessage(
-                    'hello <<[[image]]>>',
-                    image=lf_modalities.Image.from_uri('https://fake/image')
-                ),
-                sampling_options=lf.LMSamplingOptions(n=2)
-            ),
-            'Sample 0 for message: https://fake/image',
-        )
-    lm_3 = openai.Gpt35Turbo(api_key='test_key')
-    with self.assertRaisesRegex(ValueError, 'Unsupported modality'):
-      lm_3(
-          lf.UserMessage(
-              'hello <<[[image]]>>',
-              image=lf_modalities.Image.from_uri('https://fake/image')
-          ),
-      )
-
-  def test_sample_chat_completion(self):
-    with mock.patch('requests.Session.post') as mock_request:
-      mock_request.side_effect = mock_chat_completion_request
-      openai.SUPPORTED_MODELS_AND_SETTINGS['gpt-4'].update({
-          'cost_per_1k_input_tokens': 1.0,
-          'cost_per_1k_output_tokens': 1.0,
-      })
-      lm = openai.OpenAI(api_key='test_key', model='gpt-4')
-      results = lm.sample(
-          ['hello', 'bye'], sampling_options=lf.LMSamplingOptions(n=3)
-      )
-
-    self.assertEqual(len(results), 2)
-    print(results[0])
-    self.assertEqual(
-        results[0],
-        lf.LMSamplingResult(
-            [
-                lf.LMSample(
-                    lf.AIMessage(
-                        'Sample 0 for message.',
-                        score=0.0,
-                        logprobs=None,
-                        is_cached=False,
-                        usage=lf.LMSamplingUsage(
-                            prompt_tokens=33,
-                            completion_tokens=33,
-                            total_tokens=66,
-                            estimated_cost=0.2 / 3,
-                        ),
-                        tags=[lf.Message.TAG_LM_RESPONSE],
-                    ),
-                    score=0.0,
-                    logprobs=None,
-                ),
-                lf.LMSample(
-                    lf.AIMessage(
-                        'Sample 1 for message.',
-                        score=0.0,
-                        logprobs=None,
-                        is_cached=False,
-                        usage=lf.LMSamplingUsage(
-                            prompt_tokens=33,
-                            completion_tokens=33,
-                            total_tokens=66,
-                            estimated_cost=0.2 / 3,
-                        ),
-                        tags=[lf.Message.TAG_LM_RESPONSE],
-                    ),
-                    score=0.0,
-                    logprobs=None,
-                ),
-                lf.LMSample(
-                    lf.AIMessage(
-                        'Sample 2 for message.',
-                        score=0.0,
-                        logprobs=None,
-                        is_cached=False,
-                        usage=lf.LMSamplingUsage(
-                            prompt_tokens=33,
-                            completion_tokens=33,
-                            total_tokens=66,
-                            estimated_cost=0.2 / 3,
-                        ),
-                        tags=[lf.Message.TAG_LM_RESPONSE],
-                    ),
-                    score=0.0,
-                    logprobs=None,
-                ),
-            ],
-            usage=lf.LMSamplingUsage(
-                prompt_tokens=100, completion_tokens=100, total_tokens=200,
-                estimated_cost=0.2,
-            ),
-        ),
-    )
+  def test_estimate_cost(self):
     self.assertEqual(
-        results[1],
-        lf.LMSamplingResult(
-            [
-                lf.LMSample(
-                    lf.AIMessage(
-                        'Sample 0 for message.',
-                        score=0.0,
-                        logprobs=None,
-                        is_cached=False,
-                        usage=lf.LMSamplingUsage(
-                            prompt_tokens=33,
-                            completion_tokens=33,
-                            total_tokens=66,
-                            estimated_cost=0.2 / 3,
-                        ),
-                        tags=[lf.Message.TAG_LM_RESPONSE],
-                    ),
-                    score=0.0,
-                    logprobs=None,
-                ),
-                lf.LMSample(
-                    lf.AIMessage(
-                        'Sample 1 for message.',
-                        score=0.0,
-                        logprobs=None,
-                        is_cached=False,
-                        usage=lf.LMSamplingUsage(
-                            prompt_tokens=33,
-                            completion_tokens=33,
-                            total_tokens=66,
-                            estimated_cost=0.2 / 3,
-                        ),
-                        tags=[lf.Message.TAG_LM_RESPONSE],
-                    ),
-                    score=0.0,
-                    logprobs=None,
-                ),
-                lf.LMSample(
-                    lf.AIMessage(
-                        'Sample 2 for message.',
-                        score=0.0,
-                        logprobs=None,
-                        is_cached=False,
-                        usage=lf.LMSamplingUsage(
-                            prompt_tokens=33,
-                            completion_tokens=33,
-                            total_tokens=66,
-                            estimated_cost=0.2 / 3,
-                        ),
-                        tags=[lf.Message.TAG_LM_RESPONSE],
-                    ),
-                    score=0.0,
-                    logprobs=None,
-                ),
-            ],
-            usage=lf.LMSamplingUsage(
-                prompt_tokens=100, completion_tokens=100, total_tokens=200,
-                estimated_cost=0.2,
-            ),
+        openai.Gpt4(api_key='test_key').estimate_cost(
+            num_input_tokens=100, num_output_tokens=100
         ),
+        0.009
     )
 
-  def test_sample_with_contextual_options(self):
-    with mock.patch('requests.Session.post') as mock_request:
-      mock_request.side_effect = mock_chat_completion_request
-      lm = openai.OpenAI(api_key='test_key', model='text-davinci-003')
-      with lf.use_settings(sampling_options=lf.LMSamplingOptions(n=2)):
-        results = lm.sample(['hello'])
-
-    self.assertEqual(len(results), 1)
-    self.assertEqual(
-        results[0],
-        lf.LMSamplingResult(
-            [
-                lf.LMSample(
-                    lf.AIMessage(
-                        'Sample 0 for message.',
-                        score=0.0,
-                        logprobs=None,
-                        is_cached=False,
-                        usage=lf.LMSamplingUsage(
-                            prompt_tokens=50,
-                            completion_tokens=50,
-                            total_tokens=100,
-                        ),
-                        tags=[lf.Message.TAG_LM_RESPONSE],
-                    ),
-                    score=0.0,
-                    logprobs=None,
-                ),
-                lf.LMSample(
-                    lf.AIMessage(
-                        'Sample 1 for message.',
-                        score=0.0,
-                        logprobs=None,
-                        is_cached=False,
-                        usage=lf.LMSamplingUsage(
-                            prompt_tokens=50,
-                            completion_tokens=50,
-                            total_tokens=100,
-                        ),
-                        tags=[lf.Message.TAG_LM_RESPONSE],
-                    ),
-                    score=0.0,
-                    logprobs=None,
-                ),
-            ],
-            usage=lf.LMSamplingUsage(
-                prompt_tokens=100, completion_tokens=100, total_tokens=200
-            ),
-        )
-    )
-
-  def test_call_with_system_message(self):
-    with mock.patch('requests.Session.post') as mock_request:
-      mock_request.side_effect = mock_chat_completion_request
-      lm = openai.OpenAI(api_key='test_key', model='gpt-4')
-      self.assertEqual(
-          lm(
-              lf.UserMessage(
-                  'hello',
-                  system_message='hi',
-              ),
-              sampling_options=lf.LMSamplingOptions(n=2)
-          ),
-          '''Sample 0 for message. system=[{'type': 'text', 'text': 'hi'}]''',
-      )
-
-  def test_call_with_json_schema(self):
-    with mock.patch('requests.Session.post') as mock_request:
-      mock_request.side_effect = mock_chat_completion_request
-      lm = openai.OpenAI(api_key='test_key', model='gpt-4')
-      self.assertEqual(
-          lm(
-              lf.UserMessage(
-                  'hello',
-                  json_schema={
-                      'type': 'object',
-                      'properties': {
-                          'name': {'type': 'string'},
-                      },
-                      'required': ['name'],
-                      'title': 'Person',
-                  }
-              ),
-              sampling_options=lf.LMSamplingOptions(n=2)
-          ),
-          'Sample 0 for message. format=json_schema',
-      )
-
-    # Test bad json schema.
-    with self.assertRaisesRegex(ValueError, '`json_schema` must be a dict'):
-      lm(lf.UserMessage('hello', json_schema='foo'))
-
-    with self.assertRaisesRegex(
-        ValueError, 'The root of `json_schema` must have a `title` field'
-    ):
-      lm(lf.UserMessage('hello', json_schema={}))
-
 
 if __name__ == '__main__':
   unittest.main()

langfun/core/modalities/mime.py

@@ -138,9 +138,17 @@ class Mime(lf.Modality):
 
   @property
   def content_uri(self) -> str:
+    """Returns the URI with encoded content."""
     base64_content = base64.b64encode(self.to_bytes()).decode()
     return f'data:{self.mime_type};base64,{base64_content}'
 
+  @property
+  def embeddable_uri(self) -> str:
+    """Returns the URI that can be embedded in HTML."""
+    if self.uri and self.uri.lower().startswith(('http:', 'https:', 'ftp:')):
+      return self.uri
+    return self.content_uri
+
   @classmethod
   def from_uri(cls, uri: str, **kwargs) -> 'Mime':
     if cls is Mime:

langfun/core/modalities/mime_test.py

@@ -23,12 +23,12 @@ import pyglove as pg
 
 def mock_request(*args, **kwargs):
   del args, kwargs
-  return pg.Dict(content='foo')
+  return pg.Dict(content=b'foo')
 
 
 def mock_readfile(*args, **kwargs):
   del args, kwargs
-  return 'bar'
+  return b'bar'
 
 
 class CustomMimeTest(unittest.TestCase):
@@ -65,17 +65,32 @@ class CustomMimeTest(unittest.TestCase):
     ):
       mime.Custom('text/plain')
 
+  def test_uri(self):
+    content = mime.Custom.from_uri('http://mock/web/a.txt', mime='text/plain')
+    with mock.patch('requests.get') as mock_requests_stub:
+      mock_requests_stub.side_effect = mock_request
+      self.assertEqual(content.uri, 'http://mock/web/a.txt')
+      self.assertEqual(content.content_uri, 'data:text/plain;base64,Zm9v')
+      self.assertEqual(content.embeddable_uri, 'http://mock/web/a.txt')
+
+    content = mime.Custom.from_uri('a.txt', mime='text/plain')
+    with mock.patch('pyglove.io.readfile') as mock_readfile_stub:
+      mock_readfile_stub.side_effect = mock_readfile
+      self.assertEqual(content.uri, 'a.txt')
+      self.assertEqual(content.content_uri, 'data:text/plain;base64,YmFy')
+      self.assertEqual(content.embeddable_uri, 'data:text/plain;base64,YmFy')
+
   def test_from_uri(self):
     content = mime.Custom.from_uri('http://mock/web/a.txt', mime='text/plain')
     with mock.patch('requests.get') as mock_requests_stub:
       mock_requests_stub.side_effect = mock_request
-      self.assertEqual(content.to_bytes(), 'foo')
+      self.assertEqual(content.to_bytes(), b'foo')
       self.assertEqual(content.mime_type, 'text/plain')
 
     content = mime.Custom.from_uri('a.txt', mime='text/plain')
     with mock.patch('pyglove.io.readfile') as mock_readfile_stub:
       mock_readfile_stub.side_effect = mock_readfile
-      self.assertEqual(content.to_bytes(), 'bar')
+      self.assertEqual(content.to_bytes(), b'bar')
       self.assertEqual(content.mime_type, 'text/plain')
 
   def assert_html_content(self, html, expected):

langfun/core/modality_test.py

@@ -11,7 +11,6 @@
 # WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
 # See the License for the specific language governing permissions and
 # limitations under the License.
-"""Tests for modality."""
 from typing import Any
 import unittest
 

{langfun-0.1.2.dev202501090804.dist-info → langfun-0.1.2.dev202501110804.dist-info}/METADATA

@@ -1,6 +1,6 @@
 Metadata-Version: 2.2
 Name: langfun
-Version: 0.1.2.dev202501090804
+Version: 0.1.2.dev202501110804
 Summary: Langfun: Language as Functions.
 Home-page: https://github.com/google/langfun
 Author: Langfun Authors

{langfun-0.1.2.dev202501090804.dist-info → langfun-0.1.2.dev202501110804.dist-info}/RECORD

@@ -16,7 +16,7 @@ langfun/core/memory.py,sha256=f-asN1F7Vehgdn_fK84v73GrEUOxRtaW934keutTKjk,2416
 langfun/core/message.py,sha256=16oiMpg9O9VKrgpfrvJrfvga3n3FzUuD_zdWb9nvSWA,25686
 langfun/core/message_test.py,sha256=jtZoNBNbA99i2fjoKg5vTRgoUe84J4MH8ZMGakGmTHs,32577
 langfun/core/modality.py,sha256=K8pUGuMpfWcOtVcXC_OqVjro1-RhHF6ddQni61DuYzM,4166
-langfun/core/modality_test.py,sha256=7SwhixFME2Q1sIXRgJx97EZFiIyC31A9NVr6_nDtFv4,2441
+langfun/core/modality_test.py,sha256=0WL_yd3B4K-FviWdSpDnOwj0f9TQI0v9t6X0vWvvJbo,2415
 langfun/core/natural_language.py,sha256=3ynSnaYQnjE60LIPK5fyMgdIjubnPYZwzGq4rWPeloE,1177
 langfun/core/natural_language_test.py,sha256=LHGU_1ytbkGuSZQFIFP7vP3dBlcY4-A12fT6dbjUA0E,1424
 langfun/core/sampling.py,sha256=SCnS5PFJWNVxSKvSkSCNRUmruvScun8UcNN4gafuXcw,5866
@@ -75,25 +75,27 @@ langfun/core/eval/v2/reporting.py,sha256=QOp5jX761Esvi5w_UIRLDqPY_XRO6ru02-DOrdq
 langfun/core/eval/v2/reporting_test.py,sha256=UmYSAQvD3AIXsSyWQ-WD2uLtEISYpmBeoKY5u5Qwc8E,5696
 langfun/core/eval/v2/runners.py,sha256=DKEmSlGXjOXKWFdBhTpLy7tMsBHZHd1Brl3hWIngsSQ,15931
 langfun/core/eval/v2/runners_test.py,sha256=A37fKK2MvAVTiShsg_laluJzJ9AuAQn52k7HPbfD0Ks,11666
-langfun/core/llms/__init__.py,sha256=AvKRIryNNDgL3fchzNp25NPh95ojI5lOnT8XHaJS_BE,6639
+langfun/core/llms/__init__.py,sha256=Ntr0kvHc17VEZ5EV9fCoYY1kzRvQxCoZrtDRYNiMWCs,6742
 langfun/core/llms/anthropic.py,sha256=a5MmnFsBA0CbfvwzXT1v_0fqLRMrhUNdh1tx6469PQ4,14357
 langfun/core/llms/anthropic_test.py,sha256=-2U4kc_pgBM7wqxu8RuxzyHPGww1EAWqKUvN4PW8Btw,8058
 langfun/core/llms/compositional.py,sha256=csW_FLlgL-tpeyCOTVvfUQkMa_zCN5Y2I-YbSNuK27U,2872
 langfun/core/llms/compositional_test.py,sha256=4eTnOer-DncRKGaIJW2ZQQMLnt5r2R0UIx_DYOvGAQo,2027
-langfun/core/llms/deepseek.py,sha256=HxdLFFar5Mh01kKMyoLGnoYciACwxe2UzSpvKtXbSyw,8471
-langfun/core/llms/deepseek_test.py,sha256=Eo-FFyW8PWeFH-AbfY6Ib2MARnFlMtEtPIJ86DsTyz4,14856
+langfun/core/llms/deepseek.py,sha256=Y7DlLUWrukbPVyBMesppd-m75Q-PxD0b3KnMKaoY_8I,3744
+langfun/core/llms/deepseek_test.py,sha256=dS72i52bwMpCN4dJDvpJI59AnNChpwxS5eYYFrhGh90,1843
 langfun/core/llms/fake.py,sha256=gCHBYBLvBCsC78HI1hpoqXCS-p1FMTgY1P1qh_sGBPk,3070
 langfun/core/llms/fake_test.py,sha256=2h13qkwEz_JR0mtUDPxdAhQo7MueXaFSwsD2DIRDW9g,7653
 langfun/core/llms/gemini.py,sha256=tfM4vrt0WnvnrxRhWXZWh7Gp8dYYfMnSbi9uOstkSak,17399
 langfun/core/llms/gemini_test.py,sha256=2ERhYWCJwnfDTQbCaZHFuB1TdWJFrOBS7yyCBInIdQk,6129
 langfun/core/llms/google_genai.py,sha256=85Vmx5QmsziON03PRsFQINSu5NF6pAAuFFhUdDteWGc,3662
 langfun/core/llms/google_genai_test.py,sha256=JZf_cbQ4GGGpwiQCLjFJn7V4jxBBqgZhIx91AzbGKVo,1250
-langfun/core/llms/groq.py,sha256=dCnR3eAECEKuKKAAj-PDTs8NRHl6CQPdf57m1f6a79U,10312
-langfun/core/llms/groq_test.py,sha256=GYF_Qtq5S1H1TrKH38t6_lkdroqT7v-joYLDKnmS9e0,5274
-langfun/core/llms/llama_cpp.py,sha256=9tXQntSCDtjTF3bnyJrAPCr4N6wycy5nXYvp9uduygE,2843
-langfun/core/llms/llama_cpp_test.py,sha256=MWO_qaOeKjRniGjcaWPDScd7HPaIJemqUZoslrt4FPs,1806
-langfun/core/llms/openai.py,sha256=g5X_ySW-b2f0uRE8sb3_W1sAB-pWpKLNNoflBoRBwrc,21080
-langfun/core/llms/openai_test.py,sha256=kOWa1nf-nJvtYY10REUw5wojh3ZgfU8tRaCZ8wUgJbA,16623
+langfun/core/llms/groq.py,sha256=oGxyyCi5TtMVu2POdO8pXS7pK4Es56FtqjghZIGYopc,7579
+langfun/core/llms/groq_test.py,sha256=9aOD3nO4dEoH57B-5iOr5DQjG0vyv1jzFtAe7HfoiNg,1963
+langfun/core/llms/llama_cpp.py,sha256=Z7P3gc4xeIjc2bX0Ey1y5EUYJVMnMa2Q67PZ9iye9sE,1409
+langfun/core/llms/llama_cpp_test.py,sha256=wfTO7nmUwL65U2kK9P9fcMt92JjNDuVia4G1E7znf_4,1086
+langfun/core/llms/openai.py,sha256=CHUg8c0FE30LZkhvj0Kc8w0oqckbZhrp45GA94HWPs8,16614
+langfun/core/llms/openai_compatible.py,sha256=MF9JCc0uTPkIK95uvQTMIBXk4z_xKQMZqQHFaVmXwcA,5898
+langfun/core/llms/openai_compatible_test.py,sha256=TSEKhwy7eg-BW5tgRHwoprngE4j5F9HzEVuRoPK8sWw,16183
+langfun/core/llms/openai_test.py,sha256=m85YjGCvWvV5ZYagjC0FqI0FcqyCEVCbUUs8Wm3iUrc,2475
 langfun/core/llms/rest.py,sha256=sWbYUV8S3SuOg9giq7xwD-xDRfaF7NP_ig7bI52-Rj4,3442
 langfun/core/llms/rest_test.py,sha256=NZ3Nf0XQVpT9kLP5cBVo_yBHLI7vWTYhWQxYEJVMGs4,3472
 langfun/core/llms/vertexai.py,sha256=MuwLPTJ6-9x2uRDCSM1_biPK6M76FFlL1ezf5OmobDA,5504
@@ -110,8 +112,8 @@ langfun/core/modalities/audio.py,sha256=qCrVCX690SG0ps-ZfOtNWvHn_CmdJsmxF7GySScW
 langfun/core/modalities/audio_test.py,sha256=yyGEBYqMXmNs_F2dEtj-PX8HE040vqh-YQppsvdxPw4,2025
 langfun/core/modalities/image.py,sha256=ovcX8NLBNv8WzxAdhQ-u4VQfHVBkWijRj_oiNwhE9lk,1768
 langfun/core/modalities/image_test.py,sha256=XMgtJXY75R5eo0CZ222D1QUy57_hESnttmCGWwDLt7k,3824
-langfun/core/modalities/mime.py,sha256=jKE1eDaP46isD-IjWgfzhxnTcmQJ-b4wCoq6tury3J8,7576
-langfun/core/modalities/mime_test.py,sha256=lC99gsaTUwVkxQat9hkR3CV-1GI-QPMgwaIdf9EqnoM,4471
+langfun/core/modalities/mime.py,sha256=xZpnENHqwIUBLbT3s3izJvAPnw3XTtXe2v1y4mz9UkI,7856
+langfun/core/modalities/mime_test.py,sha256=Lg21nKFi--a884gnI7tFS4cHnwByEEhWg4Ugsr-UCbw,5277
 langfun/core/modalities/ms_office.py,sha256=0PnM6W9SovVfJNd4XCpUfOcGBuvj4hlA1zqPjdsJNFM,3725
 langfun/core/modalities/ms_office_test.py,sha256=rrDBx51ELAqRaUX6Y6tgg0k3tEFHOlRMfUpeULD7mlo,87868
 langfun/core/modalities/pdf.py,sha256=mfaeCbUA4JslFVTARiJh8hW7imvL4tLVw9gUhO5bAZA,727
@@ -148,8 +150,8 @@ langfun/core/templates/demonstration.py,sha256=vCrgYubdZM5Umqcgp8NUVGXgr4P_c-fik
 langfun/core/templates/demonstration_test.py,sha256=SafcDQ0WgI7pw05EmPI2S4v1t3ABKzup8jReCljHeK4,2162
 langfun/core/templates/selfplay.py,sha256=yhgrJbiYwq47TgzThmHrDQTF4nDrTI09CWGhuQPNv-s,2273
 langfun/core/templates/selfplay_test.py,sha256=Ot__1P1M8oJfoTp-M9-PQ6HUXqZKyMwvZ5f7yQ3yfyM,2326
-langfun-0.1.2.dev202501090804.dist-info/LICENSE,sha256=WNHhf_5RCaeuKWyq_K39vmp9F28LxKsB4SpomwSZ2L0,11357
-langfun-0.1.2.dev202501090804.dist-info/METADATA,sha256=mzbVRkRzSywU6qRzqZJ5KeGgTUGpYjMepZqt6OJtMJw,8172
-langfun-0.1.2.dev202501090804.dist-info/WHEEL,sha256=In9FTNxeP60KnTkGw7wk6mJPYd_dQSjEZmXdBdMCI-8,91
-langfun-0.1.2.dev202501090804.dist-info/top_level.txt,sha256=RhlEkHxs1qtzmmtWSwYoLVJAc1YrbPtxQ52uh8Z9VvY,8
-langfun-0.1.2.dev202501090804.dist-info/RECORD,,
+langfun-0.1.2.dev202501110804.dist-info/LICENSE,sha256=WNHhf_5RCaeuKWyq_K39vmp9F28LxKsB4SpomwSZ2L0,11357
+langfun-0.1.2.dev202501110804.dist-info/METADATA,sha256=U4RBtp5S0e6UddQqFSW3ywZeLG_2vJRmcFTR7m5cOCQ,8172
+langfun-0.1.2.dev202501110804.dist-info/WHEEL,sha256=In9FTNxeP60KnTkGw7wk6mJPYd_dQSjEZmXdBdMCI-8,91
+langfun-0.1.2.dev202501110804.dist-info/top_level.txt,sha256=RhlEkHxs1qtzmmtWSwYoLVJAc1YrbPtxQ52uh8Z9VvY,8
+langfun-0.1.2.dev202501110804.dist-info/RECORD,,