python-fastllm 0.0.63__tar.gz → 0.0.64__tar.gz

{python_fastllm-0.0.63 → python_fastllm-0.0.64}/PKG-INFO

@@ -1,6 +1,6 @@
 Metadata-Version: 2.4
 Name: python-fastllm
-Version: 0.0.63
+Version: 0.0.64
 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-0.0.64}/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-0.0.64}/fastllm/__init__.py

@@ -2,4 +2,4 @@
 
 - `fastllm.responses`: Provider-independent Responses resources and streaming over FastLLM"""
 
-__version__ = "0.0.63"
+__version__ = "0.0.64"

{python_fastllm-0.0.63 → python_fastllm-0.0.64}/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-0.0.64}/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))

{python_fastllm-0.0.63 → python_fastllm-0.0.64}/fastllm/anthropic.py

@@ -1,6 +1,6 @@
 """Translate Anthropic Messages API requests, responses, and streams to and from fastllm's shared message types, with tool replay, media, and prompt caching"""
 
-# AUTOGENERATED! DO NOT EDIT! File to edit: ../nbs/04_anthropic.ipynb.
+# AUTOGENERATED! DO NOT EDIT! File to edit: ../nbs/06_anthropic.ipynb.
 
 # %% auto #0
 __all__ = ['ant_tc_types', 'norm_tool_call', 'norm_tool_calls', 'norm_usage', 'finalize_usage', 'norm_finish', 'search_results',
@@ -9,7 +9,7 @@ __all__ = ['ant_tc_types', 'norm_tool_call', 'norm_tool_calls', 'norm_usage', 'f
            'denorm_tool_choice', 'denorm_reasoning', 'denorm_web_search', 'denorm_system', 'denorm_user',
            'denorm_image', 'denorm_file', 'denorm_tool_result', 'apply_cache_idxs', 'mk_payload', 'get_hdrs', 'cost']
 
-# %% ../nbs/04_anthropic.ipynb #02afd3d7
+# %% ../nbs/06_anthropic.ipynb #02afd3d7
 import json, copy
 from collections import Counter
 from fastcore.utils import *
@@ -22,7 +22,7 @@ from aidialog.msg_parts import (MdStr, Part, PartType, mk_part, Msg, Completion,
 from .streaming import *
 from .streaming import mk_acollect_stream
 
-# %% ../nbs/04_anthropic.ipynb #5656d1d6
+# %% ../nbs/06_anthropic.ipynb #5656d1d6
 ant_tc_types = ("tool_use", "server_tool_use", "mcp_tool_use")
 
 def norm_tool_call(b):
@@ -37,7 +37,7 @@ def norm_tool_calls(resp, delta=False):
         if tc:= norm_tool_call(b): out.append(tc)
     return out
 
-# %% ../nbs/04_anthropic.ipynb #0a1e2a8e
+# %% ../nbs/06_anthropic.ipynb #0a1e2a8e
 def norm_usage(resp):
     "Normalize Anthropic usage shape."
     if not (usg:=resp.get("usage")): return None
@@ -58,7 +58,7 @@ def finalize_usage(usg, parts):
     return Usage(prompt_tokens=usg.prompt_tokens, completion_tokens=ct-rt, total_tokens=usg.prompt_tokens+ct,
         cached_tokens=usg.cached_tokens, cache_creation_tokens=usg.cache_creation_tokens, reasoning_tokens=rt, search_queries=usg.search_queries, raw=usg.raw)
 
-# %% ../nbs/04_anthropic.ipynb #7a8b1f8f
+# %% ../nbs/06_anthropic.ipynb #7a8b1f8f
 def norm_finish(resp, tcs=None):
     "Canonicalize finish_reason to OpenAI Chat values: stop, tool_calls, length, content_filter."
     reason = resp.get("stop_reason")
@@ -66,7 +66,7 @@ def norm_finish(resp, tcs=None):
     r =  mp.get(reason, reason)
     return FinishReason.tool_calls if r==FinishReason.stop and any(~L(tcs).attrgot('server')) else r 
 
-# %% ../nbs/04_anthropic.ipynb #9796be28
+# %% ../nbs/06_anthropic.ipynb #9796be28
 def search_results(parts, blocks):
     "Give each server call the sources its `web_search_tool_result` block reports, one `title: url` line each, or the block's error code"
     calls = {p.id: p for p in parts if isinstance(p, ToolUse) and p.server}
@@ -76,7 +76,7 @@ def search_results(parts, blocks):
         call.text = res.get('error_code', '') if isinstance(res, dict) else MdStr('\n'.join(f"{r.get('title', '')}: {r['url']}" for r in res if r.get('url')))
     return parts
 
-# %% ../nbs/04_anthropic.ipynb #20d2e803
+# %% ../nbs/06_anthropic.ipynb #20d2e803
 def _ant_part_type(typ):
     "Map Anthropic content block type to canonical PartType."
     ant_parts_mapping = dict(text=PartType.text, thinking=PartType.thinking, redacted_thinking=PartType.thinking, tool_use=PartType.tool_use, server_tool_use=PartType.tool_use, mcp_tool_use=PartType.tool_use)
@@ -97,12 +97,12 @@ def norm_parts(resp):
             parts.append(Text(b.get("text", ""), citations=cites, raw=b))
     return search_results(parts, resp.get("content", []))
 
-# %% ../nbs/04_anthropic.ipynb #f24e52db
+# %% ../nbs/06_anthropic.ipynb #f24e52db
 def raw_msg(resp):
     "The assistant message as the API sent it"
     return dict(role='assistant', content=resp['content'])
 
-# %% ../nbs/04_anthropic.ipynb #5d1440b7
+# %% ../nbs/06_anthropic.ipynb #5d1440b7
 def norm_tr_parts(content):
     "Normalize an Anthropic `tool_result` content str or block list to canonical `Part`s"
     if isinstance(content, str): return [Text(content)]
@@ -118,7 +118,7 @@ def norm_tr_parts(content):
         else: parts.append(mk_part(**b))
     return parts
 
-# %% ../nbs/04_anthropic.ipynb #a3869e31
+# %% ../nbs/06_anthropic.ipynb #a3869e31
 def norm_sse_event(ev, **kwargs):
     ev = obj2dict(ev)
     typ = ev.get("type")
@@ -142,12 +142,12 @@ def norm_sse_event(ev, **kwargs):
     elif typ == "error": raise api_error_from_event(ev)
     return Delta(text=text, thinking=thinking, tool_calls=tcs, citations=citations, raw=ev, **kwargs)
 
-# %% ../nbs/04_anthropic.ipynb #33cd3bd3
+# %% ../nbs/06_anthropic.ipynb #33cd3bd3
 def delta_index_fn(d, typ, last_typ, last_idx):
     'Returns accumulation index for current delta and updated last idx'
     return nested_idx(d, 'raw', 'index'), None
 
-# %% ../nbs/04_anthropic.ipynb #de900e34
+# %% ../nbs/06_anthropic.ipynb #de900e34
 def collate_raw(deltas):
     "The assistant message rebuilt from a stream's events"
     blocks = {}
@@ -164,7 +164,7 @@ def collate_raw(deltas):
         elif typ == 'content_block_stop' and '_json' in blocks[i]: blocks[i]['input'] = json.loads(blocks[i].pop('_json') or '{}')
     return dict(role='assistant', content=list(blocks.values()))
 
-# %% ../nbs/04_anthropic.ipynb #c994adb1
+# %% ../nbs/06_anthropic.ipynb #c994adb1
 @delegates(mk_acollect_stream, but=['index_fn', 'api_name'])
 async def acollect_stream(resp, **kwargs):
     res = mk_acollect_stream(norm_and_yield(resp, norm_sse_event), index_fn=delta_index_fn, api_name='anthropic', **kwargs)
@@ -172,16 +172,16 @@ async def acollect_stream(resp, **kwargs):
         if isinstance(o, Completion): search_results(o.message.content, o.message.raw['content'])
         yield o
 
-# %% ../nbs/04_anthropic.ipynb #bb11ab62
+# %% ../nbs/06_anthropic.ipynb #bb11ab62
 def _ant_cc(block, p):
     "Add the part's prompt-cache directive to `block`, if it has one."
     if p.cache_control: block['cache_control'] = p.cache_control
     return block
 
-# %% ../nbs/04_anthropic.ipynb #62e9a042
+# %% ../nbs/06_anthropic.ipynb #62e9a042
 def _sanid(id_str): return re.sub(r'[^a-zA-Z0-9_-]', '_', id_str or '')
 
-# %% ../nbs/04_anthropic.ipynb #6ec772cb
+# %% ../nbs/06_anthropic.ipynb #6ec772cb
 def denorm_tool_use(p:ToolUse):
     "Convert canonical tool_use Part to Anthropic tool_use content block."
     return _ant_cc(dict(type='tool_use', id=_sanid(p.id or ''), name=p.name or '', input=p.arguments or {}), p)
@@ -209,7 +209,7 @@ def denorm_msgs(msgs:list[Msg]):
         elif m.role == 'tool':      res.append(denorm_tool(m))
     return res
 
-# %% ../nbs/04_anthropic.ipynb #f762a6f2
+# %% ../nbs/06_anthropic.ipynb #f762a6f2
 def denorm_tool_schs(tools):
     "Convert canonical tools to Anthropic format."
     out = []
@@ -220,7 +220,7 @@ def denorm_tool_schs(tools):
         out.append(dict(name=name, description=desc, input_schema=params))
     return out
 
-# %% ../nbs/04_anthropic.ipynb #6118d7ea
+# %% ../nbs/06_anthropic.ipynb #6118d7ea
 def denorm_tool_choice(v):
     "Map canonical tool_choice to Anthropic format."
     if v is None: return None
@@ -229,7 +229,7 @@ def denorm_tool_choice(v):
     if v in ('none', 'off', 'disabled'):  return {'type': 'none'}
     return {'type': 'tool', 'name': v}
 
-# %% ../nbs/04_anthropic.ipynb #310f1e75
+# %% ../nbs/06_anthropic.ipynb #310f1e75
 def denorm_reasoning(v):
     "Map canonical reasoning_effort to Anthropic adaptive thinking + output_config."
     mp = dict(minimal='low', low='low', medium='medium', high='high', xhigh='xhigh', max='max')
@@ -239,7 +239,7 @@ def denorm_reasoning(v):
     elif isinstance(v, str) and v in mp:  return {"thinking": {"type": "adaptive"}, "output_config": {"effort": mp.get(v)}}
     raise err
 
-# %% ../nbs/04_anthropic.ipynb #8fa9fbb8
+# %% ../nbs/06_anthropic.ipynb #8fa9fbb8
 def denorm_web_search(v):
     "Map canonical web_search_options to Anthropic hosted web_search tool."
     _max_uses = dict(low=1, medium=5, high=10)
@@ -249,13 +249,13 @@ def denorm_web_search(v):
     if (u := (v or {}).get("user_location", {}).get("approximate")): t["user_location"] = {"type": "approximate", **u}
     return t
 
-# %% ../nbs/04_anthropic.ipynb #6b485275
+# %% ../nbs/06_anthropic.ipynb #6b485275
 def denorm_system(sp):
     if isinstance(sp, Msg): sp = sp.content[0]
     if isinstance(sp, Part): return [_ant_cc(dict(type='text', text=sp.text), sp)]
     else: return sp
 
-# %% ../nbs/04_anthropic.ipynb #1b990cc9
+# %% ../nbs/06_anthropic.ipynb #1b990cc9
 def denorm_user(m:Msg):
     "Convert canonical user Msg to Anthropic user message."
     parts = []
@@ -267,19 +267,19 @@ def denorm_user(m:Msg):
         elif isinstance(p, InputFile):  parts.append(_ant_cc(denorm_file(p), p))
     return dict(role='user', content=parts)
 
-# %% ../nbs/04_anthropic.ipynb #edd87272
+# %% ../nbs/06_anthropic.ipynb #edd87272
 def denorm_image(p, max_sz=None):
     if (b64:=data_url(p.text)):
         data = resize_b64(b64[1], max_sz) if max_sz else b64[1]
         return {"type": "image", "source": dict(type="base64", media_type=b64[0], data=data)}
     return {"type": "image", "source": {"type": "url", "url": p.text}}
 
-# %% ../nbs/04_anthropic.ipynb #fc6bbdfc
+# %% ../nbs/06_anthropic.ipynb #fc6bbdfc
 def denorm_file(p):
     if (b64:=data_url(p.text)): return {"type": "document", "source": dict(type="base64", media_type=b64[0], data=b64[1])}
     return {"type": "document", "source": {"type": "url", "url": p.text}}
 
-# %% ../nbs/04_anthropic.ipynb #37b63cc2
+# %% ../nbs/06_anthropic.ipynb #37b63cc2
 def denorm_tool_result(p:ToolResult):
     "Convert canonical tool_result Part to Anthropic tool_result content block."
     tid = _sanid(p.id or (p.raw or {}).get('call_id',''))
@@ -293,7 +293,7 @@ def denorm_tool_result(p:ToolResult):
         return _ant_cc(dict(type='tool_result', tool_use_id=tid, content=blocks), p)
     return _ant_cc(dict(type='tool_result', tool_use_id=tid, content=tool_text(p.text)), p)
 
-# %% ../nbs/04_anthropic.ipynb #e4eb8d3b
+# %% ../nbs/06_anthropic.ipynb #e4eb8d3b
 def _add_cache_control(
     msg,     # Canonical Msg to mark
     ttl=None # Cache TTL: '5m' (default) or '1h'
@@ -304,7 +304,7 @@ def _add_cache_control(
     if idxs: msg.content[idxs[-1]].cache_control = cc
     return msg
 
-# %% ../nbs/04_anthropic.ipynb #6bc52ef5
+# %% ../nbs/06_anthropic.ipynb #6bc52ef5
 def _clear_cache_control(obj):
     "Clear canonical cache marks from a message or part."
     if isinstance(obj, Msg):
@@ -332,7 +332,7 @@ def apply_cache_idxs(
         except IndexError: continue
     return sp
 
-# %% ../nbs/04_anthropic.ipynb #587d4fe5
+# %% ../nbs/06_anthropic.ipynb #587d4fe5
 @delegates(payload_kwargs)
 def mk_payload(msgs, model, **kwargs):
     sp = kwargs.get('system')
@@ -347,7 +347,7 @@ def mk_payload(msgs, model, **kwargs):
     if (temp:=kwargs.get('temperature')) is not None: payload['temperature'] = temp
     return payload
 
-# %% ../nbs/04_anthropic.ipynb #60d52c1f
+# %% ../nbs/06_anthropic.ipynb #60d52c1f
 def get_hdrs(api_key=None, oauth_token=None):
     "An OAuth token travels as a bearer under Anthropic's OAuth beta; an API key as `x-api-key`"
     hdrs = {"anthropic-version": "2023-06-01"}
@@ -355,7 +355,7 @@ def get_hdrs(api_key=None, oauth_token=None):
     return hdrs | {"x-api-key": get_api_key(api_key, 'ANTHROPIC_API_KEY')}
 
 
-# %% ../nbs/04_anthropic.ipynb #0d03643a
+# %% ../nbs/06_anthropic.ipynb #0d03643a
 def cost(usage, m):
     raw = usage.raw
     in_tok = raw['input_tokens']
@@ -370,7 +370,7 @@ def cost(usage, m):
     cost += cache_1h   * m.get('cache_creation_input_token_cost_above_1hr', 0)
     return cost
 
-# %% ../nbs/04_anthropic.ipynb #f7c0b989
+# %% ../nbs/06_anthropic.ipynb #f7c0b989
 api_registry.register('anthropic', norm_tool_calls=norm_tool_calls, norm_parts=norm_parts, norm_finish=norm_finish, norm_usage=norm_usage,
     finalize_usage=finalize_usage, raw_msg=raw_msg, collate_raw=collate_raw, acollect_stream=acollect_stream, mk_payload=mk_payload,
     cost=cost, get_hdrs=get_hdrs, endpoint='/v1/messages')