python-fastllm 0.0.4__py3-none-any.whl → 0.0.5__py3-none-any.whl

fastllm/__init__.py

@@ -1 +1 @@
-__version__ = "0.0.4"
+__version__ = "0.0.5"

fastllm/acomplete.py

@@ -62,7 +62,7 @@ def mk_client(model, vendor_name=None, api_name=None, api_key=None, base_url=Non
                 if auth_fn.exists(): api_key = nested_idx(json.loads(auth_fn.read_text()), *keys)
             api_key = get_api_key(api_key, env_api_nm)
         except KeyError: raise ValueError(f"Unknown vendor '{vendor_name}', {err_msg}")
-    elif api_name and base_url and api_key:  vendor_name = ifnone(vendor_name, 'custom')
+    elif base_url and api_key: vendor_name, api_name = ifnone(vendor_name, 'custom'), ifnone(api_name, 'openai_chat')
     elif (api_name:=infer_api_name(model)):  base_url, vendor_name = None, api_name
     else: raise ValueError(f"Model {model} can't be auto resolved, {err_msg}")
     api = api_registry.apis[api_name]

fastllm/chat.py

@@ -476,7 +476,7 @@ def _think_kw(model, think, vendor_name):
 def _prep_call(self:AsyncChat, prefill, search, max_tokens, kwargs, stream=False, think=None):
     "Prepare model info, prefill, search, and provider kwargs for a completion call"
     model_info = get_model_info(self.model, self.vendor_name)
-    if max_tokens is None: max_tokens = model_info.get('max_output_tokens')
+    if max_tokens is None: max_tokens = ifnone(model_info.get('max_output_tokens'), 32_000)
     if not model_info.get("supports_assistant_prefill"): prefill = None
     if _has_search(model_info) and (s:=ifnone(search,self.search)):
         if 'web_search_options' not in kwargs: kwargs['web_search_options'] = {}

fastllm/types.py

@@ -240,7 +240,7 @@ def infer_api_name(model):
 def get_model_meta(model, vendor_name=None, tfm=noop):
     "Look up cost metadata for `model` from litellm price map, using `vendor_name` prefix if needed."
     vendor_name = ifnone(vendor_name, infer_api_name(model))
-    mp = model_prices_meta()
+    mp, key = model_prices_meta(), ''
     if model in mp: key = model
     elif vendor_name=='gemini' and model.startswith('models/'): key = f"gemini/{model.removeprefix('models/')}"
     elif vendor_name:                                           key = f"{vendor_name}/{model}"
@@ -273,7 +273,7 @@ _codex_overrides = {
 }
 
 # %% ../nbs/00_types.ipynb #fbfdeb0a
-def get_model_info(mn, vendor_name=None):
+def get_model_info(mn, vendor_name=None, strict=False):
     info = get_model_meta(mn, 'chatgpt' if vendor_name=='codex' else vendor_name)
     # anthropic web search 
     if 'search_context_cost_per_query' in info: info['supports_web_search'] = True
@@ -306,6 +306,9 @@ def get_model_info(mn, vendor_name=None):
                     supports_prompt_caching=True, supports_native_streaming=True, supports_native_structured_output=True,
                     max_tokens=1000000, max_input_tokens=1000000, max_output_tokens=65536,
                     input_cost_per_token=0.5e-6, cache_read_input_token_cost=0.1e-6, output_cost_per_token=3.0e-6)
+    
+    # unresolved models
+    if not info and not strict: info = info | codex_pricing
     return dict2obj(info)
 
 # %% ../nbs/00_types.ipynb #8bfca02d

{python_fastllm-0.0.4.dist-info → python_fastllm-0.0.5.dist-info}/METADATA

@@ -1,6 +1,6 @@
 Metadata-Version: 2.4
 Name: python-fastllm
-Version: 0.0.4
+Version: 0.0.5
 Author-email: Kerem Turgutlu <keremturgutlu@gmail.com>
 License: Apache-2.0
 Project-URL: Repository, https://github.com/AnswerDotAI/fastllm

{python_fastllm-0.0.4.dist-info → python_fastllm-0.0.5.dist-info}/RECORD

@@ -1,21 +1,21 @@
-fastllm/__init__.py,sha256=1mptEzQihbdyqqzMgdns_j5ZGK9gz7hR2bsgA_TnjO4,22
+fastllm/__init__.py,sha256=S7u1lbuWmM3A3ajykBialmPoJUK6Jg-WmNqM-9OZFdk,22
 fastllm/_modidx.py,sha256=oWXxw9eciYsnR2YAhla6c7HCPdXSFb9jV7WQ71fZXPQ,29551
-fastllm/acomplete.py,sha256=DTKUDx4Ed6_NdFlUj2XdQCY47p5tO0yv9ALB_22XVks,6615
+fastllm/acomplete.py,sha256=juAS2UutHSNJYA2OgisKGc5l991yg3dCn5QLNIgqAUY,6644
 fastllm/anthropic.py,sha256=fG20kOv3d3wGKQe8rD5pFWgZHKe-vT-9QJ3nPXh2twY,14615
-fastllm/chat.py,sha256=iVPHMbHFlUL8bIiUlYp4GqPhsCGn0pcl9M9LnKwj6lg,32921
+fastllm/chat.py,sha256=3zDfqcMt_v8NNDiOyhNz8czd4vlOzA7r0_GL1zNt-uI,32937
 fastllm/gemini.py,sha256=E1EYMfV8IMpC_-WzlDrkhz_CJQmzmxvaVUucNgPOqSA,14947
 fastllm/openai_chat.py,sha256=wZ0HI0m9ipy9XVhqmYBXf-BmkVAOipUVwqu9NGB_rJU,10941
 fastllm/openai_responses.py,sha256=Nk5bfTCF2-a17nwvIsf-u39j539v9KIduVfScECItKk,13052
 fastllm/streaming.py,sha256=FYG4-rt7mfGKrUPxCMp_Z3kkxGLKKAVPev8ifD3YFlQ,6652
-fastllm/types.py,sha256=2Tuf4AicfCSsf4AbkYWthPqyarGPru-LAkcPEWxQeLc,12192
+fastllm/types.py,sha256=JQdMrAdkqYOIeDtuNC18CFtsXjp12FqTp2_3OlOroV8,12304
 fastllm/specs/anthropic.json,sha256=VCgTjM2_HoDpCkeu3q_TCOEZLMHriJZLAG3LnDBAgGM,541035
 fastllm/specs/anthropic.yml,sha256=3S3NAKdXB1Nwp-Sn9Gmh4tBnwhGGhMO3DXkGqPXPUYs,724122
 fastllm/specs/gemini.json,sha256=zJGOdvZ2BvCiTENZt0-BDEvNBMl8h6EBmEskle_WBto,309331
 fastllm/specs/openai.with-code-samples.json,sha256=Kto19AW1u8MfxVDJ4cFVBIdZQOIyy8NWylswo57eABU,1995929
 fastllm/specs/openai.with-code-samples.yml,sha256=DlcWGdaeP4k7smVjt6UbyehJ-2XGU3rn3nCIBMDRfYU,2553630
 fastllm/specs/spec_manifest.json,sha256=9tVFwojXFnNqsAxQzCRTP1lgSIM0fXixnrXdv4Cmb0c,653
-python_fastllm-0.0.4.dist-info/METADATA,sha256=akEn90MvhkmJUpjbL8k28h9bDeOIbxOIUDmbDAIFqws,19546
-python_fastllm-0.0.4.dist-info/WHEEL,sha256=aeYiig01lYGDzBgS8HxWXOg3uV61G9ijOsup-k9o1sk,91
-python_fastllm-0.0.4.dist-info/entry_points.txt,sha256=dq0chsiRjJYStCOXweFW9L6LpyMTjWu2AabKCbTSbuI,36
-python_fastllm-0.0.4.dist-info/top_level.txt,sha256=F8qodL7nEGUHGmzzqfhNKCTIr1i0D6cvudOnm3z7o0Y,8
-python_fastllm-0.0.4.dist-info/RECORD,,
+python_fastllm-0.0.5.dist-info/METADATA,sha256=j24Nj_C1FR9tUVG3loB_xMkc_05qOAbZ9JmimlK6fPc,19546
+python_fastllm-0.0.5.dist-info/WHEEL,sha256=aeYiig01lYGDzBgS8HxWXOg3uV61G9ijOsup-k9o1sk,91
+python_fastllm-0.0.5.dist-info/entry_points.txt,sha256=dq0chsiRjJYStCOXweFW9L6LpyMTjWu2AabKCbTSbuI,36
+python_fastllm-0.0.5.dist-info/top_level.txt,sha256=F8qodL7nEGUHGmzzqfhNKCTIr1i0D6cvudOnm3z7o0Y,8
+python_fastllm-0.0.5.dist-info/RECORD,,