python-fastllm 0.0.60__tar.gz → 0.0.62__tar.gz

{python_fastllm-0.0.60 → python_fastllm-0.0.62}/PKG-INFO

@@ -1,6 +1,6 @@
 Metadata-Version: 2.4
 Name: python-fastllm
-Version: 0.0.60
+Version: 0.0.62
 Author-email: Kerem Turgutlu <keremturgutlu@gmail.com>
 License: Apache-2.0
 Project-URL: Repository, https://github.com/AnswerDotAI/fastllm
@@ -61,15 +61,11 @@ 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', {}),
-    ('accounts/fireworks/models/kimi-k3', dict(vendor_name='fireworks_ai'))
-]
+models = [('claude-sonnet-5', {}), ('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.")],
-                       model=name, max_tokens=mtok, **kw)
+        model=name, max_tokens=mtok, **kw)
     text = ''.join(p.text for p in r.message.content if isinstance(p, Text))
     print(f"{name}: {text.strip()}")
 ```
@@ -101,11 +97,8 @@ r = await stream([user("What should I do today?")], model='models/gemini-3-flash
 Define tools in the shared schema. `fastllm` translates it to the provider’s tool format. This example handles a tool request, adds the result to the history, and continues the conversation:
 
 ``` python
-tools = [{"type": "function", "function": {
-    "name": "get_weather",
-    "description": "Get current weather for a city",
-    "parameters": {"type": "object", "properties": {"city": {"type": "string"}}, "required": ["city"]}
-}}]
+tools = [{"type": "function", "function": dict(name="get_weather", description="Get current weather for a city",
+    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)
@@ -118,7 +111,7 @@ 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.")]
+    user("Should I bring a jacket? Answer in one short sentence.")]
 r2 = await stream(msgs, model='claude-sonnet-5', tools=tools, max_tokens=mtok)
 ```
 
@@ -129,12 +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='claude-sonnet-5', 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='claude-sonnet-5', tools=tools, tool_choice='none', max_tokens=mtok)
 print("None:", r.tool_calls)
 ```
 
@@ -148,14 +139,13 @@ 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',
-                 reasoning_effort='low', max_tokens=8192)
+    reasoning_effort='low', max_tokens=8192)
 for p in r.message.content:
     if isinstance(p, Thinking): print(f"Thinking excerpt: {p.text[:150]}...")
 
 print("Kimi: ", end='')
-r = await stream([user("What is 127 × 849? Return just the number.")],
-                 model='accounts/fireworks/models/kimi-k3', vendor_name='fireworks_ai',
-                 reasoning_effort='low', max_tokens=8192)
+r = await stream([user('What is 127 × 849? Return just the number.')], model='accounts/fireworks/models/kimi-k3',
+    vendor_name='fireworks_ai', reasoning_effort='low', max_tokens=8192)
 for p in r.message.content:
     if isinstance(p, Thinking): print(f"Thinking excerpt: {p.text[:150]}...")
 ```
@@ -172,7 +162,7 @@ OpenAI’s Responses API supports server-side web search. Server tool calls are
 ``` python
 ws_tools = [{"type": "web_search_preview"}]
 r = await acomplete([user("What is the latest Python release? Reply with the version number only.")],
-                    model='gpt-4o-mini', tools=ws_tools, max_tokens=512)
+    model='gpt-4o-mini', tools=ws_tools, max_tokens=512)
 print("Server tools used:", [tc.name for tc in r.tool_calls if tc.server])
 ```
 
@@ -187,11 +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='claude-sonnet-5', 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='claude-sonnet-5', system=system, max_tokens=mtok)
 print("Second call: cache read tokens", r2.usage.cached_tokens)
 ```
 
@@ -203,10 +193,7 @@ print("Second call: cache read tokens", r2.usage.cached_tokens)
 Use `InputImage` with a model that supports images. The example sends the same image message to three providers:
 
 ``` python
-img_msg = Msg(role='user', content=[
-    InputImage(sample_img_url),
-    Text("List three visible objects, using nouns only.")
-])
+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', {})]:
     print(f"{name}: ", end='')

{python_fastllm-0.0.60 → python_fastllm-0.0.62}/README.md

@@ -45,15 +45,11 @@ 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', {}),
-    ('accounts/fireworks/models/kimi-k3', dict(vendor_name='fireworks_ai'))
-]
+models = [('claude-sonnet-5', {}), ('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.")],
-                       model=name, max_tokens=mtok, **kw)
+        model=name, max_tokens=mtok, **kw)
     text = ''.join(p.text for p in r.message.content if isinstance(p, Text))
     print(f"{name}: {text.strip()}")
 ```
@@ -85,11 +81,8 @@ r = await stream([user("What should I do today?")], model='models/gemini-3-flash
 Define tools in the shared schema. `fastllm` translates it to the provider’s tool format. This example handles a tool request, adds the result to the history, and continues the conversation:
 
 ``` python
-tools = [{"type": "function", "function": {
-    "name": "get_weather",
-    "description": "Get current weather for a city",
-    "parameters": {"type": "object", "properties": {"city": {"type": "string"}}, "required": ["city"]}
-}}]
+tools = [{"type": "function", "function": dict(name="get_weather", description="Get current weather for a city",
+    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)
@@ -102,7 +95,7 @@ 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.")]
+    user("Should I bring a jacket? Answer in one short sentence.")]
 r2 = await stream(msgs, model='claude-sonnet-5', tools=tools, max_tokens=mtok)
 ```
 
@@ -113,12 +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='claude-sonnet-5', 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='claude-sonnet-5', tools=tools, tool_choice='none', max_tokens=mtok)
 print("None:", r.tool_calls)
 ```
 
@@ -132,14 +123,13 @@ 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',
-                 reasoning_effort='low', max_tokens=8192)
+    reasoning_effort='low', max_tokens=8192)
 for p in r.message.content:
     if isinstance(p, Thinking): print(f"Thinking excerpt: {p.text[:150]}...")
 
 print("Kimi: ", end='')
-r = await stream([user("What is 127 × 849? Return just the number.")],
-                 model='accounts/fireworks/models/kimi-k3', vendor_name='fireworks_ai',
-                 reasoning_effort='low', max_tokens=8192)
+r = await stream([user('What is 127 × 849? Return just the number.')], model='accounts/fireworks/models/kimi-k3',
+    vendor_name='fireworks_ai', reasoning_effort='low', max_tokens=8192)
 for p in r.message.content:
     if isinstance(p, Thinking): print(f"Thinking excerpt: {p.text[:150]}...")
 ```
@@ -156,7 +146,7 @@ OpenAI’s Responses API supports server-side web search. Server tool calls are
 ``` python
 ws_tools = [{"type": "web_search_preview"}]
 r = await acomplete([user("What is the latest Python release? Reply with the version number only.")],
-                    model='gpt-4o-mini', tools=ws_tools, max_tokens=512)
+    model='gpt-4o-mini', tools=ws_tools, max_tokens=512)
 print("Server tools used:", [tc.name for tc in r.tool_calls if tc.server])
 ```
 
@@ -171,11 +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='claude-sonnet-5', 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='claude-sonnet-5', system=system, max_tokens=mtok)
 print("Second call: cache read tokens", r2.usage.cached_tokens)
 ```
 
@@ -187,10 +177,7 @@ print("Second call: cache read tokens", r2.usage.cached_tokens)
 Use `InputImage` with a model that supports images. The example sends the same image message to three providers:
 
 ``` python
-img_msg = Msg(role='user', content=[
-    InputImage(sample_img_url),
-    Text("List three visible objects, using nouns only.")
-])
+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', {})]:
     print(f"{name}: ", end='')

{python_fastllm-0.0.60 → python_fastllm-0.0.62}/fastllm/__init__.py

@@ -2,4 +2,4 @@
 
 - `fastllm.responses`: Provider-independent Responses resources and streaming over FastLLM"""
 
-__version__ = "0.0.60"
+__version__ = "0.0.62"

{python_fastllm-0.0.60 → python_fastllm-0.0.62}/fastllm/acomplete.py

@@ -33,6 +33,7 @@ vendor_mapping = {
     "codex":        ('openai', 'https://chatgpt.com/backend-api/codex', 'CODEX_AUTH_TOKEN', _codex_json),
     "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"),
     "mimo":         ('anthropic', "https://api.xiaomimimo.com/anthropic", "MIMO_API_KEY"),
     "openrouter":   ('openai_chat', "https://openrouter.ai/api/v1", "OPENROUTER_API_KEY"),
     "together":     ('openai_chat', "https://api.together.xyz/v1", "TOGETHER_API_KEY"),
@@ -167,20 +168,23 @@ 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, xtra_hdrs=xtra_hdrs)
+    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)
     api = api_registry[api_name]
     if previous_response_id is not None:
         if not getattr(api, 'supports_previous_response_id', False):
             raise ValueError('previous_response_id is not supported by this transport')
         kwargs['previous_response_id'] = previous_response_id
     payload = api.mk_payload(msgs, model, stream=stream, **kwargs)
-    xtra_body = ifnone(xtra_body, {})
     if fix := getattr(api, 'fix_payload', None): fix(payload, model, vendor_name)
+    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}
         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, payload | xtra_body, None if own else cli._url(path))
-    async def _call(): return await provider_req(cli, path, payload | xtra_body, params=params, stream=stream)
+    if defaults.debug_mode: _debug_print(model, api_name, vendor_name, body, None if own else cli._url(path))
+    async def _call(): return await provider_req(cli, path, body, params=params, stream=stream, headers=headers)
     async def _mk_gen():
         resp = payload if own else await _call()  # an own-transport api's `acollect_stream` makes the request itself, from the payload
         async for o in api.acollect_stream(resp, model=model, vendor_name=vendor_name, stop_callables=stop_callables): yield o

{python_fastllm-0.0.60 → python_fastllm-0.0.62}/fastllm/chat.py

@@ -9,6 +9,7 @@ __all__ = ['effort', 'contents', 'stop_reason', 'cite_footnote', 'postproc', 'Re
            'StopSequencesCallback', 'StreamAccum', 'adisplay_stream']
 
 # %% ../nbs/07_chat.ipynb #d5a3bc1f
+import secrets
 from typing import Optional,Callable
 from fastcore.funccall import mk_ns, call_func, call_func_async, get_schema
 from fastcore.utils import *
@@ -168,6 +169,7 @@ class AsyncChat:
         cache=False,              # Anthropic prompt caching
         cache_idxs:list=[-1],     # Anthropic cache breakpoint idxs, use `0` for sys prompt if provided
         ttl=None,                 # Anthropic prompt caching ttl
+        prompt_cache_key=None,     # Stable conversation key for OpenAI Responses; generated once if omitted
         api_name=None,            # API to use, one of ApiName: openai (responses), openai_chat, anthropic, gemini
         vendor_name=None,         # Vendor name, one of vendor_mapping which resolves api_base/api_key automatically
         api_key=None,             # API key when model can't be resolved or vendor_name is not known or codex
@@ -195,6 +197,7 @@ class AsyncChat:
         self.response_id,self._response_hist_idx = None,0
         self._turn_start = 0    # index into `hist` where the current turn began, for `full`
         self.last_req_use = None  # usage of the latest request only; `use` accumulates across a turn's tool-call steps
+        if prompt_cache_key is None: prompt_cache_key = secrets.token_hex(16)
         store_attr(but='cbs')
         self.cbs = L()
         if default_cbs: self.add_cbs(defaults.chat_callbacks)
@@ -256,6 +259,7 @@ def _prep_call(self:AsyncChat, search, max_tokens, kwargs, stream=False, think=N
     if self.base_url:      kwargs['base_url'] = self.base_url
     if self.endpoint:      kwargs['endpoint'] = self.endpoint
     if self.extra_headers: kwargs['xtra_hdrs'] = self.extra_headers
+    kwargs.setdefault('prompt_cache_key', self.prompt_cache_key)
     kwargs.update(_think_kw(self.model, think, self.vendor_name))
     return max_tokens