python-fastllm 0.0.63__tar.gz → 0.0.65__tar.gz

{python_fastllm-0.0.63 → python_fastllm-0.0.65}/PKG-INFO

@@ -1,6 +1,6 @@
 Metadata-Version: 2.4
 Name: python-fastllm
-Version: 0.0.63
+Version: 0.0.65
 Author-email: Kerem Turgutlu <keremturgutlu@gmail.com>
 License-Expression: Apache-2.0
 Project-URL: Repository, https://github.com/AnswerDotAI/fastllm
@@ -32,7 +32,7 @@ Run `pip install python-fastllm` or clone from github and install locally.
 ``` python
 from aidialog.msg_parts import Msg, Part, Text, Thinking, ToolUse, InputImage, mk_tool_res_msg, Completion
 from fastllm.acomplete import acomplete
-from fastllm.types import sample_img_url, sample_doms
+from fastllm.types import sample_img_url, sample_doms, sonn, sonn46
 import asyncio, json
 ```
 
@@ -61,7 +61,7 @@ mtok = 1024
 Use the same `acomplete` interface with Claude, GPT, Gemini, or Kimi. Set `model` and any provider-specific routing arguments:
 
 ``` python
-models = [('claude-sonnet-5', {}), ('gpt-4o-mini', {}), ('models/gemini-3-flash-preview', {}),
+models = [(sonn, {}), ('gpt-4o-mini', {}), ('models/gemini-3-flash-preview', {}),
     ('accounts/fireworks/models/kimi-k3', dict(vendor_name='fireworks_ai'))]
 for name, kw in models:
     r = await acomplete([user("Translate 'hello' into French. Return only the translation.")],
@@ -83,7 +83,7 @@ Pass `system` to supply a system prompt. `fastllm` translates it to the provider
 sys = "You are a pirate chef. Always respond in pirate speak and mention food. Use one short sentence."
 
 print("Claude: ", end='')
-r = await stream([user("What should I do today?")], model='claude-sonnet-5', system=sys, max_tokens=mtok)
+r = await stream([user("What should I do today?")], model=sonn, system=sys, max_tokens=mtok)
 
 print("Gemini: ", end='')
 r = await stream([user("What should I do today?")], model='models/gemini-3-flash-preview', system=sys, max_tokens=mtok)
@@ -101,7 +101,7 @@ tools = [{"type": "function", "function": dict(name="get_weather", description="
     parameters=dict(type="object", properties={"city": {"type": "string"}}, required=["city"]))}]
 
 msgs = [user("What's the weather in Paris?")]
-r1 = await stream(msgs, model='claude-sonnet-5', tools=tools, max_tokens=mtok)
+r1 = await stream(msgs, model=sonn, tools=tools, max_tokens=mtok)
 print("Tool calls:", r1.tool_calls)
 ```
 
@@ -112,7 +112,7 @@ print("Tool calls:", r1.tool_calls)
 ``` python
 msgs += [r1.message, mk_tool_res_msg(r1.tool_calls, ['22°C, sunny with light clouds']),
     user("Should I bring a jacket? Answer in one short sentence.")]
-r2 = await stream(msgs, model='claude-sonnet-5', tools=tools, max_tokens=mtok)
+r2 = await stream(msgs, model=sonn, tools=tools, max_tokens=mtok)
 ```
 
     No, a jacket isn't necessary — it's a mild, sunny 22°C day.
@@ -122,10 +122,10 @@ r2 = await stream(msgs, model='claude-sonnet-5', tools=tools, max_tokens=mtok)
 Control whether the model must use tools, can’t use tools, or decides on its own:
 
 ``` python
-r = await acomplete([user('Hello there!')], model='claude-sonnet-5', tools=tools, tool_choice='required', max_tokens=mtok)
+r = await acomplete([user('Hello there!')], model=sonn, tools=tools, tool_choice='required', max_tokens=mtok)
 print("Required:", [tc.name for tc in r.tool_calls])
 
-r = await acomplete([user("What's the weather?")], model='claude-sonnet-5', tools=tools, tool_choice='none', max_tokens=mtok)
+r = await acomplete([user("What's the weather?")], model=sonn, tools=tools, tool_choice='none', max_tokens=mtok)
 print("None:", r.tool_calls)
 ```
 
@@ -138,7 +138,7 @@ Set `reasoning_effort` to `low`, `medium`, or `high` for supported models. The a
 
 ``` python
 print("Claude: ", end='')
-r = await stream([user("What is 127 × 849? Return just the number.")], model='claude-sonnet-4-6',
+r = await stream([user("What is 127 × 849? Return just the number.")], model=sonn46,
     reasoning_effort='low', max_tokens=8192)
 for p in r.message.content:
     if isinstance(p, Thinking): print(f"Thinking excerpt: {p.text[:150]}...")
@@ -177,11 +177,11 @@ long_ctx = "You are an expert on the solar system. " * 200
 system = Text(long_ctx, cache_control={'type': 'ephemeral'})
 
 r1 = await acomplete([user("What is Jupiter's mass? Answer in one sentence.")],
-    model='claude-sonnet-5', system=system, max_tokens=mtok)
+    model=sonn, system=system, max_tokens=mtok)
 print("First call: cache creation tokens", r1.usage.cache_creation_tokens)
 
 r2 = await acomplete([user("What is Saturn's mass? Answer in one sentence.")],
-    model='claude-sonnet-5', system=system, max_tokens=mtok)
+    model=sonn, system=system, max_tokens=mtok)
 print("Second call: cache read tokens", r2.usage.cached_tokens)
 ```
 
@@ -195,7 +195,7 @@ Use `InputImage` with a model that supports images. The example sends the same i
 ``` python
 img_msg = Msg(role='user', content=[InputImage(sample_img_url), Text('List three visible objects, using nouns only.')])
 
-for name, kw in [('claude-sonnet-5', {}), ('gpt-4o-mini', {}), ('models/gemini-3-flash-preview', {})]:
+for name, kw in [(sonn, {}), ('gpt-4o-mini', {}), ('models/gemini-3-flash-preview', {})]:
     print(f"{name}: ", end='')
     r = await stream([img_msg], model=name, max_tokens=mtok, **kw)
 ```

{python_fastllm-0.0.63 → python_fastllm-0.0.65}/README.md

@@ -16,7 +16,7 @@ Run `pip install python-fastllm` or clone from github and install locally.
 ``` python
 from aidialog.msg_parts import Msg, Part, Text, Thinking, ToolUse, InputImage, mk_tool_res_msg, Completion
 from fastllm.acomplete import acomplete
-from fastllm.types import sample_img_url, sample_doms
+from fastllm.types import sample_img_url, sample_doms, sonn, sonn46
 import asyncio, json
 ```
 
@@ -45,7 +45,7 @@ mtok = 1024
 Use the same `acomplete` interface with Claude, GPT, Gemini, or Kimi. Set `model` and any provider-specific routing arguments:
 
 ``` python
-models = [('claude-sonnet-5', {}), ('gpt-4o-mini', {}), ('models/gemini-3-flash-preview', {}),
+models = [(sonn, {}), ('gpt-4o-mini', {}), ('models/gemini-3-flash-preview', {}),
     ('accounts/fireworks/models/kimi-k3', dict(vendor_name='fireworks_ai'))]
 for name, kw in models:
     r = await acomplete([user("Translate 'hello' into French. Return only the translation.")],
@@ -67,7 +67,7 @@ Pass `system` to supply a system prompt. `fastllm` translates it to the provider
 sys = "You are a pirate chef. Always respond in pirate speak and mention food. Use one short sentence."
 
 print("Claude: ", end='')
-r = await stream([user("What should I do today?")], model='claude-sonnet-5', system=sys, max_tokens=mtok)
+r = await stream([user("What should I do today?")], model=sonn, system=sys, max_tokens=mtok)
 
 print("Gemini: ", end='')
 r = await stream([user("What should I do today?")], model='models/gemini-3-flash-preview', system=sys, max_tokens=mtok)
@@ -85,7 +85,7 @@ tools = [{"type": "function", "function": dict(name="get_weather", description="
     parameters=dict(type="object", properties={"city": {"type": "string"}}, required=["city"]))}]
 
 msgs = [user("What's the weather in Paris?")]
-r1 = await stream(msgs, model='claude-sonnet-5', tools=tools, max_tokens=mtok)
+r1 = await stream(msgs, model=sonn, tools=tools, max_tokens=mtok)
 print("Tool calls:", r1.tool_calls)
 ```
 
@@ -96,7 +96,7 @@ print("Tool calls:", r1.tool_calls)
 ``` python
 msgs += [r1.message, mk_tool_res_msg(r1.tool_calls, ['22°C, sunny with light clouds']),
     user("Should I bring a jacket? Answer in one short sentence.")]
-r2 = await stream(msgs, model='claude-sonnet-5', tools=tools, max_tokens=mtok)
+r2 = await stream(msgs, model=sonn, tools=tools, max_tokens=mtok)
 ```
 
     No, a jacket isn't necessary — it's a mild, sunny 22°C day.
@@ -106,10 +106,10 @@ r2 = await stream(msgs, model='claude-sonnet-5', tools=tools, max_tokens=mtok)
 Control whether the model must use tools, can’t use tools, or decides on its own:
 
 ``` python
-r = await acomplete([user('Hello there!')], model='claude-sonnet-5', tools=tools, tool_choice='required', max_tokens=mtok)
+r = await acomplete([user('Hello there!')], model=sonn, tools=tools, tool_choice='required', max_tokens=mtok)
 print("Required:", [tc.name for tc in r.tool_calls])
 
-r = await acomplete([user("What's the weather?")], model='claude-sonnet-5', tools=tools, tool_choice='none', max_tokens=mtok)
+r = await acomplete([user("What's the weather?")], model=sonn, tools=tools, tool_choice='none', max_tokens=mtok)
 print("None:", r.tool_calls)
 ```
 
@@ -122,7 +122,7 @@ Set `reasoning_effort` to `low`, `medium`, or `high` for supported models. The a
 
 ``` python
 print("Claude: ", end='')
-r = await stream([user("What is 127 × 849? Return just the number.")], model='claude-sonnet-4-6',
+r = await stream([user("What is 127 × 849? Return just the number.")], model=sonn46,
     reasoning_effort='low', max_tokens=8192)
 for p in r.message.content:
     if isinstance(p, Thinking): print(f"Thinking excerpt: {p.text[:150]}...")
@@ -161,11 +161,11 @@ long_ctx = "You are an expert on the solar system. " * 200
 system = Text(long_ctx, cache_control={'type': 'ephemeral'})
 
 r1 = await acomplete([user("What is Jupiter's mass? Answer in one sentence.")],
-    model='claude-sonnet-5', system=system, max_tokens=mtok)
+    model=sonn, system=system, max_tokens=mtok)
 print("First call: cache creation tokens", r1.usage.cache_creation_tokens)
 
 r2 = await acomplete([user("What is Saturn's mass? Answer in one sentence.")],
-    model='claude-sonnet-5', system=system, max_tokens=mtok)
+    model=sonn, system=system, max_tokens=mtok)
 print("Second call: cache read tokens", r2.usage.cached_tokens)
 ```
 
@@ -179,7 +179,7 @@ Use `InputImage` with a model that supports images. The example sends the same i
 ``` python
 img_msg = Msg(role='user', content=[InputImage(sample_img_url), Text('List three visible objects, using nouns only.')])
 
-for name, kw in [('claude-sonnet-5', {}), ('gpt-4o-mini', {}), ('models/gemini-3-flash-preview', {})]:
+for name, kw in [(sonn, {}), ('gpt-4o-mini', {}), ('models/gemini-3-flash-preview', {})]:
     print(f"{name}: ", end='')
     r = await stream([img_msg], model=name, max_tokens=mtok, **kw)
 ```

{python_fastllm-0.0.63 → python_fastllm-0.0.65}/fastllm/__init__.py

@@ -2,4 +2,4 @@
 
 - `fastllm.responses`: Provider-independent Responses resources and streaming over FastLLM"""
 
-__version__ = "0.0.63"
+__version__ = "0.0.65"

{python_fastllm-0.0.63 → python_fastllm-0.0.65}/fastllm/_modidx.py

@@ -10,7 +10,6 @@ d = { 'settings': { 'branch': 'main',
                                    'fastllm.acomplete._classify_error': ('acomplete.html#_classify_error', 'fastllm/acomplete.py'),
                                    'fastllm.acomplete._classify_error_stream': ( 'acomplete.html#_classify_error_stream',
                                                                                  'fastllm/acomplete.py'),
-                                   'fastllm.acomplete._codex_mtime_policy': ('acomplete.html#_codex_mtime_policy', 'fastllm/acomplete.py'),
                                    'fastllm.acomplete._debug_print': ('acomplete.html#_debug_print', 'fastllm/acomplete.py'),
                                    'fastllm.acomplete._is_ctx_exceeded': ('acomplete.html#_is_ctx_exceeded', 'fastllm/acomplete.py'),
                                    'fastllm.acomplete._raise_if_done': ('acomplete.html#_raise_if_done', 'fastllm/acomplete.py'),
@@ -112,7 +111,6 @@ d = { 'settings': { 'branch': 'main',
                               'fastllm.chat._inject_tool_reminder': ('chat.html#_inject_tool_reminder', 'fastllm/chat.py'),
                               'fastllm.chat._lite_call_func': ('chat.html#_lite_call_func', 'fastllm/chat.py'),
                               'fastllm.chat._mk_tool_result': ('chat.html#_mk_tool_result', 'fastllm/chat.py'),
-                              'fastllm.chat._think_kw': ('chat.html#_think_kw', 'fastllm/chat.py'),
                               'fastllm.chat._usrtools': ('chat.html#_usrtools', 'fastllm/chat.py'),
                               'fastllm.chat.add_warning': ('chat.html#add_warning', 'fastllm/chat.py'),
                               'fastllm.chat.adisplay_stream': ('chat.html#adisplay_stream', 'fastllm/chat.py'),
@@ -126,7 +124,9 @@ d = { 'settings': { 'branch': 'main',
                               'fastllm.chat.stop_reason': ('chat.html#stop_reason', 'fastllm/chat.py'),
                               'fastllm.chat.structured': ('chat.html#structured', 'fastllm/chat.py'),
                               'fastllm.chat.think_sep': ('chat.html#think_sep', 'fastllm/chat.py')},
-            'fastllm.codex': { 'fastllm.codex._merge_tokens': ('codex.html#_merge_tokens', 'fastllm/codex.py'),
+            'fastllm.codex': { 'fastllm.codex._codex_hdrs': ('codex.html#_codex_hdrs', 'fastllm/codex.py'),
+                               'fastllm.codex._codex_vendor_auth': ('codex.html#_codex_vendor_auth', 'fastllm/codex.py'),
+                               'fastllm.codex._merge_tokens': ('codex.html#_merge_tokens', 'fastllm/codex.py'),
                                'fastllm.codex._post': ('codex.html#_post', 'fastllm/codex.py'),
                                'fastllm.codex._token_grant': ('codex.html#_token_grant', 'fastllm/codex.py'),
                                'fastllm.codex.codex_auth': ('codex.html#codex_auth', 'fastllm/codex.py'),
@@ -373,10 +373,11 @@ d = { 'settings': { 'branch': 'main',
                                'fastllm.types.Usage.__add__': ('types.html#usage.__add__', 'fastllm/types.py'),
                                'fastllm.types.Usage.__eq__': ('types.html#usage.__eq__', 'fastllm/types.py'),
                                'fastllm.types.Usage.__init__': ('types.html#usage.__init__', 'fastllm/types.py'),
-                               'fastllm.types._flat_rates': ('types.html#_flat_rates', 'fastllm/types.py'),
                                'fastllm.types.approx_pricing': ('types.html#approx_pricing', 'fastllm/types.py'),
                                'fastllm.types.effort_code': ('types.html#effort_code', 'fastllm/types.py'),
+                               'fastllm.types.effort_kwargs': ('types.html#effort_kwargs', 'fastllm/types.py'),
                                'fastllm.types.effort_levels': ('types.html#effort_levels', 'fastllm/types.py'),
+                               'fastllm.types.find_models': ('types.html#find_models', 'fastllm/types.py'),
                                'fastllm.types.fn_schema': ('types.html#fn_schema', 'fastllm/types.py'),
                                'fastllm.types.get_api_key': ('types.html#get_api_key', 'fastllm/types.py'),
                                'fastllm.types.get_model_info': ('types.html#get_model_info', 'fastllm/types.py'),
@@ -390,6 +391,7 @@ d = { 'settings': { 'branch': 'main',
                                'fastllm.types.price_tier': ('types.html#price_tier', 'fastllm/types.py'),
                                'fastllm.types.provider_req': ('types.html#provider_req', 'fastllm/types.py'),
                                'fastllm.types.register_model_info': ('types.html#register_model_info', 'fastllm/types.py'),
+                               'fastllm.types.register_sub_model': ('types.html#register_sub_model', 'fastllm/types.py'),
                                'fastllm.types.resize_b64': ('types.html#resize_b64', 'fastllm/types.py'),
                                'fastllm.types.resolve_effort': ('types.html#resolve_effort', 'fastllm/types.py'),
                                'fastllm.types.tier_rate': ('types.html#tier_rate', 'fastllm/types.py'),

{python_fastllm-0.0.63 → python_fastllm-0.0.65}/fastllm/acomplete.py

@@ -1,11 +1,11 @@
 """A unified async API for model completions and streaming"""
 
-# AUTOGENERATED! DO NOT EDIT! File to edit: ../nbs/06_acomplete.ipynb.
+# AUTOGENERATED! DO NOT EDIT! File to edit: ../nbs/03_acomplete.ipynb.
 
 # %% auto #0
 __all__ = ['vendor_mapping', 'defaults', 'split_vendor', 'mk_client', 'ContextWindowExceededError', 'acomplete']
 
-# %% ../nbs/06_acomplete.ipynb #f2f57253
+# %% ../nbs/03_acomplete.ipynb #f2f57253
 import asyncio,json,httpx2
 from fastcore.utils import *
 from fastcore.meta import *
@@ -13,7 +13,7 @@ from fasttransport.core import AsyncHttpCli
 from fasttransport.errors import APIError
 
 from .types import *
-from .codex import codex_auth_path
+from .codex import *
 from aidialog.msg_parts import (Part, PartType, Msg, Text, Thinking, ToolUse, ToolResult, InputImage, InputAudio, InputVideo,
     InputFile, mk_tool_res_msg, Completion)
 from .streaming import *
@@ -22,15 +22,13 @@ from .openai_chat import *
 from .anthropic import *
 from .gemini import *
 
-# %% ../nbs/06_acomplete.ipynb #32ee2546
-_codex_path = codex_auth_path()
-_codex_json = _codex_path, ('tokens','access_token')
+# %% ../nbs/03_acomplete.ipynb #32ee2546
 vendor_mapping = {
     "openai":       ('openai', 'https://api.openai.com/v1', 'OPENAI_API_KEY'),
     "anthropic":    ('anthropic', 'https://api.anthropic.com', 'ANTHROPIC_API_KEY'),
     "gemini":       ('gemini', 'https://generativelanguage.googleapis.com/', 'GEMINI_API_KEY'),
     "openai_chat":  ('openai_chat', 'https://api.openai.com/v1', 'OPENAI_API_KEY'),
-    "codex":        ('openai', 'https://chatgpt.com/backend-api/codex', 'CODEX_AUTH_TOKEN', _codex_json),
+    "codex":        ('openai', 'https://chatgpt.com/backend-api/codex', 'CODEX_AUTH_TOKEN'),
     "moonshot":     ('openai_chat', "https://api.moonshot.ai/v1", "MOONSHOT_API_KEY"),
     "deepseek":     ('openai_chat', "https://api.deepseek.com/v1", "DEEPSEEK_API_KEY"),
     "inception":    ('openai_chat', "https://api.inceptionlabs.ai/v1", "INCEPTION_API_KEY"),
@@ -43,7 +41,7 @@ vendor_mapping = {
     "meta":         ('openai_chat', "https://api.meta.ai/v1", "META_API_KEY"),
     "zai":          ('openai_chat', "https://api.z.ai/api/paas/v4", "ZAI_API_KEY")}
 
-# %% ../nbs/06_acomplete.ipynb #e3ed40fb
+# %% ../nbs/03_acomplete.ipynb #e3ed40fb
 def split_vendor(model):
     "Split a `'vendor/model'` path into `(vendor_name, model)`: the prefix must name a known vendor or a registered api (installed `fastllm.apis` plugins load on demand), else `(None, model)`."
     if model and '/' in model:
@@ -51,23 +49,8 @@ def split_vendor(model):
         if v in vendor_mapping or v in api_registry: return v, m
     return None, model
 
-# %% ../nbs/06_acomplete.ipynb #977a115b
-def _codex_mtime_policy():
-    "`flexicache` policy: expire codex clients when the codex auth file changes"
-    def policy(state, args, kwargs):
-        v = kwargs.get('vendor_name') or (args[1] if len(args)>1 else None)
-        if not v:
-            m = kwargs.get('model') or (args[0] if args else None)
-            v = split_vendor(m)[0] if m else None
-        if v != 'codex': return None
-        try: mt = Path(_codex_path).expanduser().stat().st_mtime
-        except OSError: mt = None
-        return mt if state != mt else None
-    policy.needs_args = True
-    return policy
-
-# %% ../nbs/06_acomplete.ipynb #79075d95
-@flexicache(_codex_mtime_policy())
+# %% ../nbs/03_acomplete.ipynb #79075d95
+@flexicache()
 def mk_client(model=None, vendor_name=None, api_name=None, api_key=None, oauth_token=None, base_url=None, xtra_hdrs=None,
     timeout=httpx2.Timeout(connect=30, read=300, write=30, pool=10)):
     err_msg = f"please pass a valid one vendor: {', '.join(list(vendor_mapping))} or pass `api_name`,`base_url` and `api_key`"
@@ -77,12 +60,8 @@ def mk_client(model=None, vendor_name=None, api_name=None, api_key=None, oauth_t
     if vendor_name:
         override_base_url = base_url
         try:
-            api_name, base_url, env_api_nm, *auth_json = vendor_mapping[vendor_name]
+            api_name, base_url, env_api_nm = vendor_mapping[vendor_name]
             base_url = override_base_url or base_url
-            if auth_json and not api_key and not oauth_token and not os.getenv(env_api_nm):
-                fn,keys = auth_json[0]  # pyright: ignore[reportAssignmentType]
-                auth_fn = Path(fn).expanduser()
-                if auth_fn.exists(): api_key = nested_idx(json.loads(auth_fn.read_text()), *keys)
             if not oauth_token: api_key = get_api_key(api_key, env_api_nm)
         except KeyError: raise ValueError(f"Unknown vendor '{vendor_name}', {err_msg}")
     elif base_url and (api_key or oauth_token): vendor_name, api_name = ifnone(vendor_name, 'custom'), ifnone(api_name, 'openai_chat')
@@ -93,7 +72,7 @@ def mk_client(model=None, vendor_name=None, api_name=None, api_key=None, oauth_t
     cli = AsyncHttpCli(base_url, base_headers=hdrs, timeout=timeout)
     return cli, api_name, vendor_name
 
-# %% ../nbs/06_acomplete.ipynb #df851a5c
+# %% ../nbs/03_acomplete.ipynb #df851a5c
 class ContextWindowExceededError(APIError): pass
 
 def _is_ctx_exceeded(code, msg):
@@ -119,8 +98,8 @@ async def _classify_error_stream(gen):
         async for x in gen: yield x
     except APIError as e: raise _classify_error(e) from e
 
-# %% ../nbs/06_acomplete.ipynb #f626a4e1
-defaults = SimpleNamespace(debug_mode=False)
+# %% ../nbs/03_acomplete.ipynb #f626a4e1
+defaults = SimpleNamespace(debug_mode=False, reasoning_effort=None)
 
 def _debug_print(model, api_name, vendor_name, payload, url):
     "Pretty-print acomplete inputs when defaults.debug_mode is set"
@@ -133,7 +112,7 @@ def _debug_print(model, api_name, vendor_name, payload, url):
     print(f"\033[1;33mpayload:\033[0m\n{pformat(p, width=120, sort_dicts=False)}")
     print('━'*60)
 
-# %% ../nbs/06_acomplete.ipynb #497c8565
+# %% ../nbs/03_acomplete.ipynb #497c8565
 async def _raise_if_done(e, n, retries, retry_delay, yielded=False):
     e = _classify_error(e)
     if yielded or not e.retryable or n == retries: raise e
@@ -154,7 +133,7 @@ async def _retry_stream(mk_gen, retries=2, retry_delay=0.5):
             return
         except APIError as e: await _raise_if_done(e, n, retries, retry_delay, yielded=yielded)
 
-# %% ../nbs/06_acomplete.ipynb #2379ec94
+# %% ../nbs/03_acomplete.ipynb #2379ec94
 @delegates(payload_kwargs)
 async def acomplete(msgs, model, api_name=None, vendor_name=None, api_key=None, oauth_token=None, base_url=None, endpoint=None, xtra_body=None, xtra_hdrs=None,
     stream=False, previous_response_id=None, stop_callables=None, retries=2, retry_delay=0.5, **kwargs):
@@ -168,7 +147,12 @@ async def acomplete(msgs, model, api_name=None, vendor_name=None, api_key=None,
     if own:
         vendor_name = ifnone(vendor_name, api_name)  # an own-transport api: no HTTP client to build
         if oauth_token: kwargs['oauth_token'] = oauth_token  # its `mk_payload` applies the user's token itself
-    else: cli, api_name, vendor_name = mk_client(model=model, vendor_name=vendor_name, api_name=api_name, api_key=api_key, oauth_token=oauth_token, base_url=base_url)
+    else:
+        auth = vendor_auth.get(vendor_name)
+        if auth and not (api_key or oauth_token or os.getenv(vendor_mapping[vendor_name][2])): api_key = await auth()
+        cli, api_name, vendor_name = mk_client(model=model, vendor_name=vendor_name, api_name=api_name, api_key=api_key, oauth_token=oauth_token, base_url=base_url)
+    if (eff := defaults.reasoning_effort) and 'reasoning_effort' not in kwargs and effort_levels(get_model_info(model, vendor_name)):
+        kwargs |= effort_kwargs(model, eff, vendor_name)
     api = api_registry[api_name]
     if previous_response_id is not None:
         if not getattr(api, 'supports_previous_response_id', False):
@@ -179,8 +163,7 @@ async def acomplete(msgs, model, api_name=None, vendor_name=None, api_key=None,
     body = payload | ifnone(xtra_body, {})
     if not own:
         headers = xtra_hdrs or {}
-        if vendor_name == 'codex' and (key := body.get('prompt_cache_key')):
-            if not any(k.lower() == 'session-id' for k in headers): headers = {**headers, 'session-id':key}
+        if hk := vendor_hdrs.get(vendor_name): headers = hk(body, headers)
         ep = endpoint or api.endpoint
         path,params = ep(model, stream) if callable(ep) else (ep, None)
     if defaults.debug_mode: _debug_print(model, api_name, vendor_name, body, None if own else cli._url(path))