python-fastllm 0.0.64__tar.gz → 0.0.65__tar.gz

{python_fastllm-0.0.64 → python_fastllm-0.0.65}/PKG-INFO

@@ -1,6 +1,6 @@
 Metadata-Version: 2.4
 Name: python-fastllm
-Version: 0.0.64
+Version: 0.0.65
 Author-email: Kerem Turgutlu <keremturgutlu@gmail.com>
 License-Expression: Apache-2.0
 Project-URL: Repository, https://github.com/AnswerDotAI/fastllm

{python_fastllm-0.0.64 → python_fastllm-0.0.65}/fastllm/__init__.py

@@ -2,4 +2,4 @@
 
 - `fastllm.responses`: Provider-independent Responses resources and streaming over FastLLM"""
 
-__version__ = "0.0.64"
+__version__ = "0.0.65"

{python_fastllm-0.0.64 → python_fastllm-0.0.65}/fastllm/anthropic.py

@@ -11,7 +11,6 @@ __all__ = ['ant_tc_types', 'norm_tool_call', 'norm_tool_calls', 'norm_usage', 'f
 
 # %% ../nbs/06_anthropic.ipynb #02afd3d7
 import json, copy
-from collections import Counter
 from fastcore.utils import *
 from fastcore.meta import *
 from fasttransport.errors import api_error_from_event
@@ -337,7 +336,9 @@ def apply_cache_idxs(
 def mk_payload(msgs, model, **kwargs):
     sp = kwargs.get('system')
     if (ci:=kwargs.get('cache_idxs')) is not None: sp = apply_cache_idxs(msgs, ci, sp, kwargs.get('ttl'))
-    payload = dict(model=model, messages=denorm_msgs(msgs), max_tokens=kwargs.get('max_tokens') or 1024)
+    max_tokens = kwargs.get('max_tokens')
+    if max_tokens is None: max_tokens = get_model_info(model, 'anthropic').get('max_output_tokens') or 1024
+    payload = dict(model=model, messages=denorm_msgs(msgs), max_tokens=max_tokens)
     if kwargs.get('stream'):                payload['stream'] = True
     if sp:                              payload['system'] = denorm_system(sp)
     if tools:=kwargs.get('tools'):          payload['tools'] = denorm_tool_schs(tools)
@@ -358,17 +359,13 @@ def get_hdrs(api_key=None, oauth_token=None):
 # %% ../nbs/06_anthropic.ipynb #0d03643a
 def cost(usage, m):
     raw = usage.raw
-    in_tok = raw['input_tokens']
-    cache_read = raw.get('cache_read_input_tokens', 0)
     cc = raw.get('cache_creation', {}) or {}
-    cache_5m  = cc.get('ephemeral_5m_input_tokens', raw.get('cache_creation_input_tokens', 0))
-    cache_1h  = cc.get('ephemeral_1h_input_tokens', 0)
-    cost  = in_tok     * m.input_cost_per_token
-    cost += raw['output_tokens'] * m.output_cost_per_token
-    cost += cache_read * m.get('cache_read_input_token_cost', 0)
-    cost += cache_5m   * m.get('cache_creation_input_token_cost', 0)
-    cost += cache_1h   * m.get('cache_creation_input_token_cost_above_1hr', 0)
-    return cost
+    prompt = dict(input_cost_per_token=raw['input_tokens'], cache_read_input_token_cost=raw.get('cache_read_input_tokens', 0),
+        cache_creation_input_token_cost=cc.get('ephemeral_5m_input_tokens', raw.get('cache_creation_input_tokens', 0)),
+        cache_creation_input_token_cost_above_1hr=cc.get('ephemeral_1h_input_tokens', 0))
+    tier = price_tier(m, sum(prompt.values()))
+    toks = prompt | dict(output_cost_per_token=raw['output_tokens'])
+    return sum(n * tier_rate(m, k, tier) for k,n in toks.items())
 
 # %% ../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,