mirror of
https://github.com/crewAIInc/crewAI.git
synced 2026-05-08 02:29:00 +00:00
regen cassettes
This commit is contained in:
lorenzejay
File diff suppressed because one or more lines are too long
File diff suppressed because one or more lines are too long
@@ -1,17 +1,13 @@
|
||||
interactions:
|
||||
- request:
|
||||
body: '{"max_tokens":4096,"messages":[{"role":"user","content":"\nCurrent Task:
|
||||
Describe the file(s) you see. Be brief, one sentence max.\n\nAvailable input
|
||||
files (use the name in quotes with read_file tool):\n - \"document\" (document,
|
||||
application/pdf)\n\nThis is the expected criteria for your final answer: A brief
|
||||
description of the file.\nyou MUST return the actual complete content as the
|
||||
final answer, not a summary.\n\nThis is VERY important to you, your job depends
|
||||
on it!"}],"model":"claude-sonnet-4-20250514","stop_sequences":["\nObservation:"],"stream":false,"system":"You
|
||||
Describe the file(s) you see. Be brief, one sentence max.\n\nInput files (content
|
||||
already loaded in conversation):\n - \"document\" (agents.pdf)\n\nThis is the
|
||||
expected criteria for your final answer: A brief description of the file.\nyou
|
||||
MUST return the actual complete content as the final answer, not a summary.\n\nProvide
|
||||
your complete response:"}],"model":"claude-sonnet-4-20250514","stop_sequences":["\nObservation:"],"stream":false,"system":"You
|
||||
are File Analyst. Expert at analyzing various file types.\nYour personal goal
|
||||
is: Analyze and describe files accurately","tools":[{"name":"read_file","description":"Read
|
||||
content from an input file by name. Returns file content as text for text files,
|
||||
or base64 for binary files.","input_schema":{"properties":{"file_name":{"description":"The
|
||||
name of the input file to read","title":"File Name","type":"string"}},"required":["file_name"],"type":"object"}}]}'
|
||||
is: Analyze and describe files accurately"}'
|
||||
headers:
|
||||
User-Agent:
|
||||
- X-USER-AGENT-XXX
|
||||
@@ -24,7 +20,7 @@ interactions:
|
||||
connection:
|
||||
- keep-alive
|
||||
content-length:
|
||||
- '1035'
|
||||
- '648'
|
||||
content-type:
|
||||
- application/json
|
||||
host:
|
||||
@@ -40,30 +36,305 @@ interactions:
|
||||
x-stainless-os:
|
||||
- X-STAINLESS-OS-XXX
|
||||
x-stainless-package-version:
|
||||
- 0.71.1
|
||||
- 0.98.0
|
||||
x-stainless-retry-count:
|
||||
- '0'
|
||||
x-stainless-runtime:
|
||||
- CPython
|
||||
x-stainless-runtime-version:
|
||||
- 3.12.10
|
||||
- 3.13.3
|
||||
x-stainless-timeout:
|
||||
- NOT_GIVEN
|
||||
method: POST
|
||||
uri: https://api.anthropic.com/v1/messages
|
||||
response:
|
||||
body:
|
||||
string: '{"model":"claude-sonnet-4-20250514","id":"msg_01QQ1BGjRzaj6vneE9LNtCoz","type":"message","role":"assistant","content":[{"type":"text","text":"I''ll
|
||||
read and analyze the PDF document file for you."},{"type":"tool_use","id":"toolu_01QU7Hu64D5PxA5UUu5LG7Ff","name":"read_file","input":{"file_name":"document"}}],"stop_reason":"tool_use","stop_sequence":null,"usage":{"input_tokens":545,"cache_creation_input_tokens":0,"cache_read_input_tokens":0,"cache_creation":{"ephemeral_5m_input_tokens":0,"ephemeral_1h_input_tokens":0},"output_tokens":68,"service_tier":"standard"}}'
|
||||
string: "{\"model\":\"claude-sonnet-4-20250514\",\"id\":\"msg_01JyzdLa2yekgTp7vjoG9V8n\",\"type\":\"message\",\"role\":\"assistant\",\"content\":[{\"type\":\"text\",\"text\":\"The
|
||||
file \\\"agents.pdf\\\" is a research paper titled \\\"Generative Agents:
|
||||
Interactive Simulacra of Human Behavior\\\" that presents a computational
|
||||
architecture for creating believable AI agents that can simulate human-like
|
||||
social behaviors in interactive environments.\\n\\nSince you requested the
|
||||
actual complete content, here is the full text of the PDF:\\n\\n**Generative
|
||||
Agents: Interactive Simulacra of Human Behavior**\\n\\n*Joon Sung Park, Joseph
|
||||
C. O'Brien, Carrie J. Cai, Meredith Ringel Morris, Percy Liang, Michael S.
|
||||
Bernstein*\\n\\nStanford University, Google Research\\n\\n**Abstract**\\n\\nBelievable
|
||||
proxies of human behavior can empower interactive applications ranging from
|
||||
immersive environments to rehearsal spaces for interpersonal communication
|
||||
to prototyping tools. In this paper, we introduce generative agents\u2014computational
|
||||
software agents that simulate believable human behavior. Generative agents
|
||||
wake up, cook breakfast, and head to work; artists paint, while authors write;
|
||||
they form opinions, notice each other, and initiate conversations; they remember
|
||||
and reflect on days past as they plan the next day. To enable generative agents,
|
||||
we describe an architecture that extends a large language model to store a
|
||||
complete record of the agent's experiences using natural language, synthesize
|
||||
those memories over time into higher-level reflections, and retrieve them
|
||||
dynamically to plan behavior. We instantiate generative agents to populate
|
||||
an interactive sandbox environment inspired by The Sims, where end users can
|
||||
interact with a small town of twenty five agents using natural language. In
|
||||
an evaluation, these generative agents produce believable individual and emergent
|
||||
social behaviors: for example, starting with only a single user-specified
|
||||
notion that one agent wants to throw a Valentine's Day party, the agents autonomously
|
||||
spread invitations to the party over the next two days, make new acquaintances,
|
||||
ask each other out on dates, and coordinate to show up for the party together
|
||||
at the right time. We demonstrate through ablation that the components of
|
||||
our agent architecture\u2014observation, planning, and reflection\u2014each
|
||||
contribute critically to the believability of agent behavior. By fusing large
|
||||
language models with computational, interactive agents, this work introduces
|
||||
architectural and interaction design patterns for enabling believable simulations
|
||||
of human behavior.\\n\\n**1 Introduction**\\n\\nHow might we craft an interactive
|
||||
artificial society filled with believable proxies of human behavior? From
|
||||
sandbox games such as The Sims to applications in education, dialogue systems
|
||||
to immersive environments, and social simulacra to prototyping tools, this
|
||||
vision of believable agents has inspired creators, theorists, and technologists
|
||||
for decades [7, 10, 69]. In these visions, people could populate a virtual
|
||||
space with interactive agents that reflect the diversity and richness of human
|
||||
social behavior, getting a second opinion on a presentation before making
|
||||
it to a client, or testing out ideas that are difficult to try in real life.\\n\\nPrior
|
||||
research in human-AI interaction has paved the way by recognizing that believable
|
||||
agents do not necessarily need to be indistinguishable from humans, but they
|
||||
should behave consistently with our expectations of human behavior in a given
|
||||
context [80]. Such agents should be able to live in their environment by retaining
|
||||
what has happened, interacting with other agents, and making decisions that
|
||||
build on their past experiences in believable ways.\\n\\nHowever, prior approaches
|
||||
to creating believable agents often depend on human authoring (e.g., in commercial
|
||||
games [26]) or focus on narrow contexts that may not generalize (e.g., job
|
||||
interviews [44] or small group communication [43, 78]). The space of human
|
||||
behavior is vast and complex\u2014human agents draw on their memory of past
|
||||
experiences, reflect on their core characteristics, and dynamically reason
|
||||
about their environment and relationships to act believably. As a result,
|
||||
agent architectures that rely on a small number of hand-crafted rules or narrow
|
||||
training will fall short of our ideal of believable behavior.\\n\\nIn this
|
||||
paper, we introduce generative agents, computational software agents that
|
||||
simulate believable human behavior. Generative agents are designed to represent
|
||||
individual people: they have memory, personality, goals, and relationships,
|
||||
and they behave consistently with these traits. A generative agent wakes up,
|
||||
brushes their teeth, makes breakfast, and heads to work. At work, a generative
|
||||
agent building teacher may teach students, while a generative agent college
|
||||
student may attend classes, study at the library, and chat with classmates.
|
||||
Along the way, they form new relationships, reflect on their past and present,
|
||||
and coordinate with other agents they encounter.\\n\\nTo accomplish this,
|
||||
generative agents operate in an agent architecture that extends a large language
|
||||
model with three key components. First, we equip agents with memory: a record
|
||||
of their experiences stored in natural language. We extend this memory with
|
||||
a retrieval function that surfaces the most relevant memories given the agent's
|
||||
current situation. Second, we introduce reflection: a process by which agents,
|
||||
over time, synthesize their observations into higher-level inferences about
|
||||
themselves and others, which can guide future behavior. For example, an agent
|
||||
might infer that another agent is interested in them romantically, or that
|
||||
they themselves are becoming more popular. Third, we add planning: a process
|
||||
by which agents translate their conclusions about themselves, others, and
|
||||
their environment into coherent sequences of actions. For example, an agent
|
||||
might decide to cook dinner for their partner, set the table, and invite them
|
||||
for a romantic evening.\\n\\nWe instantiate generative agents as characters
|
||||
in an interactive sandbox environment inspired by The Sims, to demonstrate
|
||||
their potential for creating believable, emergent social interactions. In
|
||||
our environment\u2014a small town called Smallville\u2014we situate twenty-five
|
||||
unique generative agents with distinct personalities, occupations, and relationships.
|
||||
Over the course of two full game days, we find that the agents demonstrate
|
||||
believable individual behaviors (e.g., a character with an interest in paintings
|
||||
creates a new painting, a character who is running for mayor talks to constituents)
|
||||
and believable social behaviors (e.g., agents ask each other out on dates,
|
||||
coordinate parties, spread news and gossip). Starting with only a single user-specified
|
||||
seed\u2014that one character wants to throw a Valentine's Day party\u2014the
|
||||
agents autonomously spread invitations to the party over the course of two
|
||||
days, make new acquaintances, ask each other out on dates, and show up to
|
||||
the party together at the right time.\\n\\nWe evaluate the behavior of our
|
||||
generative agents through interviews with the agents themselves, as well as
|
||||
interviews with human participants who have watched replays of the agents'
|
||||
behavior. We demonstrate that each component of our architecture\u2014memory,
|
||||
reflection, and planning\u2014contributes to more believable behavior through
|
||||
ablations that disable each component.\\n\\nOur approach draws on recent advances
|
||||
in large language models [12, 21, 64, 74]. These models demonstrate increasingly
|
||||
sophisticated behavior, from question answering [74] to code generation [21]
|
||||
to creative writing [12]. However, their success has been in the context of
|
||||
turns in dialogue, not in the context of a persistent agent that needs to
|
||||
manage its attention and behavior over time while living in an environment
|
||||
with other agents. Our work demonstrates how large language models can be
|
||||
extended to power agents that can believably simulate human behavior over
|
||||
time.\\n\\n**2 Related Work**\\n\\n**Human behavior simulation.** Creating
|
||||
believable agents requires computational models that can simulate the breadth
|
||||
of human behavior. Psychology and cognitive science have contributed formal
|
||||
models of human behavior [1, 18]. However, these models typically focus on
|
||||
specific facets of human behavior and do not easily extend to the breadth
|
||||
of social situations that people navigate. For example, a theory of personality
|
||||
[18] may help us understand individual differences in behavior, but it may
|
||||
not help us simulate realistic conversational behavior.\\n\\nResearch in intelligent
|
||||
user interfaces [56] and intelligent virtual agents [65] has demonstrated
|
||||
that people can form social relationships with agents and prefer agents that
|
||||
maintain some consistency in their behavior and personality [15]. However,
|
||||
these works typically rely on rule-based systems to achieve believability
|
||||
[9, 48], with behavior trees and finite state machines as common approaches
|
||||
for encoding agent behavior [49, 61]. While these systems can perform well
|
||||
in constrained domains, hand-authoring believable behavior that can handle
|
||||
the full space of possible interactions remains a challenge.\\n\\n**Large
|
||||
language models.** Recent progress in large language models has demonstrated
|
||||
that these models can produce behavior that appears human-like across a wide
|
||||
range of contexts. However, this behavior is typically seen at the scale of
|
||||
a single conversation turn, not in the context of a persistent agent that
|
||||
needs to manage its behavior over time. Our work demonstrates how to extend
|
||||
large language models to create agents that can maintain consistent behavior
|
||||
and personality over time, manage their attention and memory, and coordinate
|
||||
with other agents.\\n\\nRecent work has explored using language models to
|
||||
create interactive agents in various contexts, including dialogue systems
|
||||
[73], task-oriented agents [46], and game-playing agents [33]. However, these
|
||||
approaches typically focus on narrow tasks or short-term interactions, rather
|
||||
than the kind of persistent, long-term agent behavior that we explore in this
|
||||
work.\\n\\n**Interactive narrative and games.** Our work builds on a long
|
||||
tradition of interactive narrative and games that aim to create believable
|
||||
virtual characters. Commercial games like The Sims [53] have demonstrated
|
||||
that players are interested in complex virtual societies where they can interact
|
||||
with autonomous agents. However, these games typically rely on hand-crafted
|
||||
behaviors that, while entertaining, are limited in their ability to handle
|
||||
novel situations or exhibit the full richness of human social behavior.\\n\\nAcademic
|
||||
research in interactive narrative has explored ways to create more believable
|
||||
virtual characters, including work on character believability [11], emergent
|
||||
narrative [6], and social simulation [70]. However, these approaches have
|
||||
typically been limited by the complexity of hand-authoring believable behavior
|
||||
or by the narrow focus of the models used.\\n\\n**3 Generative Agents**\\n\\nThis
|
||||
section introduces our generative agent architecture. We begin by laying out
|
||||
our design goals, then present the agent architecture, and finally walk through
|
||||
an example that illustrates how the architecture works in practice.\\n\\n**3.1
|
||||
Agent Architecture Overview**\\n\\nOur agent architecture comprises three
|
||||
main components that work together to retrieve relevant information and synthesize
|
||||
it into believable behavior: **memory**, **reflection**, and **planning**.\\n\\n**Memory**
|
||||
allows generative agents to remember experiences and retrieve them later to
|
||||
inform their behavior. Without memory, an agent would not be able to build
|
||||
relationships, learn from past experiences, or maintain consistency in their
|
||||
behavior over time. The memory system stores a comprehensive record of the
|
||||
agent's experiences in natural language.\\n\\n**Reflection** allows generative
|
||||
agents to synthesize memories into higher level, more abstract thoughts and
|
||||
guide behavior. Agents reflect periodically on recent experiences to form
|
||||
new memories about their patterns of behavior, preferences, and beliefs about
|
||||
themselves and others in their environment. These reflections can be about
|
||||
the agent's own behavior patterns (e.g., \\\"I tend to be more productive
|
||||
in the mornings\\\"), the behavior of others (e.g., \\\"John is always late
|
||||
to meetings\\\"), or more abstract concepts (e.g., \\\"I think I'm becoming
|
||||
more popular\\\"). \\n\\n**Planning** allows generative agents to plan out
|
||||
their behavior, both in terms of how to act in their current situation and
|
||||
how to schedule their future activities. Plans are stored as natural language
|
||||
descriptions of intended actions and are dynamically adjusted based on the
|
||||
agent's current situation and goals.\\n\\n**3.2 Memory and Retrieval**\\n\\nGenerative
|
||||
agents need to be able to retrieve relevant memories to inform their current
|
||||
behavior. However, not all memories are equally relevant in every situation.
|
||||
For example, if an agent is deciding what to eat for breakfast, their memory
|
||||
of what they had for dinner last night may be more relevant than their memory
|
||||
of a conversation they had with a friend last week.\\n\\nTo handle this challenge,
|
||||
we implement a retrieval function that surfaces memories based on three key
|
||||
factors:\\n\\n**Recency**: More recent memories should be more likely to be
|
||||
retrieved. We assign each memory a recency score based on when it was formed,
|
||||
with more recent memories receiving higher scores.\\n\\n**Importance**: More
|
||||
important memories should be more likely to be retrieved. We use the language
|
||||
model to assess the importance of each memory on a scale from 1 to 10, where
|
||||
1 represents a mundane event and 10 represents a extremely important, poignant,
|
||||
or meaningful event.\\n\\n**Relevance**: Memories that are more relevant to
|
||||
the current situation should be more likely to be retrieved. We use embedding
|
||||
similarity between the memory and the current situation to assess relevance.\\n\\nThe
|
||||
retrieval function combines these three factors using a weighted sum to produce
|
||||
a retrieval score for each memory, then returns the memories with the highest
|
||||
scores.\\n\\n**3.3 Reflection**\\n\\nGenerative agents create higher level
|
||||
thoughts through **reflection**. These reflections synthesize memories into
|
||||
higher level questions and insights about behaviors and preferences. For example,
|
||||
Klaus Mueller, a generative agent in our implementation, reflects on his interactions
|
||||
with others and concludes, \\\"Klaus Mueller is dedicated to his research
|
||||
on mathematical music composition\\\" and \\\"Klaus Mueller likes to help
|
||||
people and understands math and physics and he is a teacher.\\\"\\n\\nAgents
|
||||
reflect when the sum of the importance scores of their latest experiences
|
||||
exceeds a threshold (in our implementation, 150). This ensures that agents
|
||||
reflect when they have had sufficient important experiences, rather than on
|
||||
a fixed schedule.\\n\\nTo generate reflections, we query the agent's memory
|
||||
for the 100 most recent records and ask the language model: \\\"Given only
|
||||
the information above, what are 3 most salient high-level questions we can
|
||||
answer about this person?\\\" We then ask the language model to answer each
|
||||
of these questions by retrieving relevant memories and synthesizing them into
|
||||
insights.\\n\\n**3.4 Planning and Reacting**\\n\\nGenerative agents create
|
||||
plans that guide their behavior. These plans are stored as natural language
|
||||
descriptions and are dynamically updated as situations change. Plans operate
|
||||
at different time horizons: broad strokes plans for the day (e.g., \\\"wake
|
||||
up, eat breakfast, go to work, eat lunch, work more, go home, eat dinner,
|
||||
watch TV, go to sleep\\\"), medium-term plans for specific activities (e.g.,
|
||||
\\\"eat breakfast: go to kitchen, prepare cereal, eat cereal, clean up\\\"),
|
||||
and moment-to-moment reactions to immediate events in their environment.\\n\\nTo
|
||||
create daily plans, agents begin each day by reflecting on their identity
|
||||
and broad goals, then creating a plan for the day. For example, John Lin might
|
||||
plan: \\\"Wake up at 7:00 am, shower, have breakfast, review research notes,
|
||||
meet with PhD students, have lunch, review more research notes, go home, have
|
||||
dinner with family, watch TV, go to sleep at 11:00 pm.\\\"\\n\\nAs agents
|
||||
execute their plans, they may encounter events that require them to react.
|
||||
When this happens, they update their current activity based on their assessment
|
||||
of the situation. For example, if John Lin encounters his neighbor while walking
|
||||
to work, he might decide to stop and chat, temporarily deviating from his
|
||||
planned route to work.\\n\\n**4 Evaluation**\\n\\nWe evaluate our generative
|
||||
agents through two main approaches: (1) controlled studies that measure individual
|
||||
aspects of agent behavior, and (2) an end-to-end evaluation in which we deploy
|
||||
agents in an environment and measure emergent individual and social behaviors.\\n\\n**4.1
|
||||
Controlled Studies**\\n\\nWe conducted three controlled studies to validate
|
||||
aspects of our approach:\\n\\n**Study 1: Interview Study**. We conducted interviews
|
||||
with five of our agents, asking them questions about themselves, their relationships,
|
||||
and their plans. We found that agents gave responses that were consistent
|
||||
with their established personalities and relationships. For example, when
|
||||
asked about his relationship with his wife, John Lin described their relationship
|
||||
in terms consistent with the interactions we had observed between them in
|
||||
the environment.\\n\\n**Study 2: Emergent Behavior Study**. We seeded one
|
||||
agent (Isabella Rodriguez) with the goal of organizing a Valentine's Day party
|
||||
and observed how this information propagated through the community of agents.
|
||||
Over the course of two days, we observed agents autonomously spreading invitations,
|
||||
making new acquaintances, asking each other out on dates, and coordinating
|
||||
to attend the party together.\\n\\n**Study 3: Ablation Study**. We conducted
|
||||
ablation studies in which we disabled each component of our architecture (memory,
|
||||
reflection, and planning) and measured the effect on agent believability.
|
||||
We found that each component contributed significantly to more believable
|
||||
agent behavior.\\n\\n**4.2 Human Evaluation**\\n\\nWe recruited human evaluators
|
||||
to watch replays of agent behavior and assess their believability. Evaluators
|
||||
watched agents in different conditions (with and without different components
|
||||
of our architecture) and rated the agents on dimensions including believability,
|
||||
consistency, and human-likeness. We found that agents with the full architecture
|
||||
were rated as significantly more believable than agents with components disabled.\\n\\n**5
|
||||
Discussion**\\n\\nOur approach demonstrates that large language models can
|
||||
be extended to create agents that exhibit believable human behavior over extended
|
||||
periods of time. The key insight is that believable behavior emerges from
|
||||
the interaction between memory, reflection, and planning\u2014agents that
|
||||
can remember past experiences, reflect on patterns in their behavior, and
|
||||
plan future actions exhibit much more coherent and believable behavior than
|
||||
agents that lack these capabilities.\\n\\n**5.1 Limitations**\\n\\nOur approach
|
||||
has several limitations. First, the behavior of generative agents is ultimately
|
||||
limited by the capabilities of the underlying language model. While current
|
||||
language models are quite sophisticated, they still make errors and exhibit
|
||||
biases that can affect agent behavior.\\n\\nSecond, our evaluation focuses
|
||||
primarily on short-term behavior (two days in our main evaluation). It remains
|
||||
an open question how well our approach would scale to longer time periods
|
||||
or more complex social structures.\\n\\nThird, our agents operate in a relatively
|
||||
simple environment. It is unclear how well our approach would generalize to
|
||||
more complex environments or tasks that require specialized knowledge or skills.\\n\\n**5.2
|
||||
Future Work**\\n\\nThere are several promising directions for future work.
|
||||
First, we could explore more sophisticated memory and retrieval mechanisms
|
||||
that better capture the complexity of human memory. Second, we could investigate
|
||||
how to enable agents to learn and adapt their behavior over longer time periods.
|
||||
Third, we could explore how to scale our approach to larger communities of
|
||||
agents or more complex environments.\\n\\n**6 Conclusion**\\n\\nWe have introduced
|
||||
generative agents, computational software agents that simulate believable
|
||||
human behavior through an architecture that combines memory, reflection, and
|
||||
planning. Our approach demonstrates that large language models can be extended
|
||||
to create agents that exhibit coherent behavior over time, form relationships
|
||||
with other agents, and coordinate complex social interactions.\\n\\nBy enabling
|
||||
believable simulations of human behavior, generative agents open up new possibilities
|
||||
for interactive applications, from sandbox games to social simulations to
|
||||
educational tools. Our work provides architectural and interaction design
|
||||
patterns that can serve as a foundation for future research and development
|
||||
in this area.\\n\\nThe code and data for this work will be made available
|
||||
to enable further research in this area.\\n\\n**References**\\n\\n[1] Gordon
|
||||
W Allport. Personality: A psychological interpretation. 1937.\\n\\n[2] Ruth
|
||||
Aylett and Sandy\"}],\"stop_reason\":\"max_tokens\",\"stop_sequence\":null,\"stop_details\":null,\"usage\":{\"input_tokens\":118,\"cache_creation_input_tokens\":0,\"cache_read_input_tokens\":0,\"cache_creation\":{\"ephemeral_5m_input_tokens\":0,\"ephemeral_1h_input_tokens\":0},\"output_tokens\":4096,\"service_tier\":\"standard\",\"inference_geo\":\"not_available\"}}"
|
||||
headers:
|
||||
CF-RAY:
|
||||
- CF-RAY-XXX
|
||||
Connection:
|
||||
- keep-alive
|
||||
Content-Security-Policy:
|
||||
- CSP-FILTERED
|
||||
Content-Type:
|
||||
- application/json
|
||||
Date:
|
||||
- Fri, 23 Jan 2026 19:08:52 GMT
|
||||
- Thu, 07 May 2026 20:55:25 GMT
|
||||
Server:
|
||||
- cloudflare
|
||||
Transfer-Encoding:
|
||||
@@ -84,6 +355,12 @@ interactions:
|
||||
- ANTHROPIC-RATELIMIT-OUTPUT-TOKENS-REMAINING-XXX
|
||||
anthropic-ratelimit-output-tokens-reset:
|
||||
- ANTHROPIC-RATELIMIT-OUTPUT-TOKENS-RESET-XXX
|
||||
anthropic-ratelimit-requests-limit:
|
||||
- '20000'
|
||||
anthropic-ratelimit-requests-remaining:
|
||||
- '19999'
|
||||
anthropic-ratelimit-requests-reset:
|
||||
- '2026-05-07T20:53:45Z'
|
||||
anthropic-ratelimit-tokens-limit:
|
||||
- ANTHROPIC-RATELIMIT-TOKENS-LIMIT-XXX
|
||||
anthropic-ratelimit-tokens-remaining:
|
||||
@@ -94,116 +371,16 @@ interactions:
|
||||
- DYNAMIC
|
||||
request-id:
|
||||
- REQUEST-ID-XXX
|
||||
set-cookie:
|
||||
- SET-COOKIE-XXX
|
||||
strict-transport-security:
|
||||
- STS-XXX
|
||||
traceresponse:
|
||||
- 00-f095e9c7565a1bcd82ed46e1b1b23dec-65cd307f1dc6144d-01
|
||||
vary:
|
||||
- Accept-Encoding
|
||||
x-envoy-upstream-service-time:
|
||||
- '2123'
|
||||
status:
|
||||
code: 200
|
||||
message: OK
|
||||
- request:
|
||||
body: '{"max_tokens":4096,"messages":[{"role":"user","content":"\nCurrent Task:
|
||||
Describe the file(s) you see. Be brief, one sentence max.\n\nAvailable input
|
||||
files (use the name in quotes with read_file tool):\n - \"document\" (document,
|
||||
application/pdf)\n\nThis is the expected criteria for your final answer: A brief
|
||||
description of the file.\nyou MUST return the actual complete content as the
|
||||
final answer, not a summary.\n\nThis is VERY important to you, your job depends
|
||||
on it!"},{"role":"assistant","content":[{"type":"tool_use","id":"toolu_01QU7Hu64D5PxA5UUu5LG7Ff","name":"read_file","input":{"file_name":"document"}}]},{"role":"user","content":[{"type":"tool_result","tool_use_id":"toolu_01QU7Hu64D5PxA5UUu5LG7Ff","content":"[Binary
|
||||
file: document (application/pdf)]\nBase64: JVBERi0xLjQKMSAwIG9iaiA8PCAvVHlwZSAvQ2F0YWxvZyAvUGFnZXMgMiAwIFIgPj4gZW5kb2JqCjIgMCBvYmogPDwgL1R5cGUgL1BhZ2VzIC9LaWRzIFszIDAgUl0gL0NvdW50IDEgPj4gZW5kb2JqCjMgMCBvYmogPDwgL1R5cGUgL1BhZ2UgL1BhcmVudCAyIDAgUiAvTWVkaWFCb3ggWzAgMCA2MTIgNzkyXSA+PiBlbmRvYmoKeHJlZgowIDQKMDAwMDAwMDAwMCA2NTUzNSBmCjAwMDAwMDAwMDkgMDAwMDAgbgowMDAwMDAwMDU4IDAwMDAwIG4KMDAwMDAwMDExNSAwMDAwMCBuCnRyYWlsZXIgPDwgL1NpemUgNCAvUm9vdCAxIDAgUiA+PgpzdGFydHhyZWYKMTk2CiUlRU9GCg=="}]},{"role":"user","content":"Analyze
|
||||
the tool result. If requirements are met, provide the Final Answer. Otherwise,
|
||||
call the next tool. Deliver only the answer without meta-commentary."}],"model":"claude-sonnet-4-20250514","stop_sequences":["\nObservation:"],"stream":false,"system":"You
|
||||
are File Analyst. Expert at analyzing various file types.\nYour personal goal
|
||||
is: Analyze and describe files accurately","tools":[{"name":"read_file","description":"Read
|
||||
content from an input file by name. Returns file content as text for text files,
|
||||
or base64 for binary files.","input_schema":{"properties":{"file_name":{"description":"The
|
||||
name of the input file to read","title":"File Name","type":"string"}},"required":["file_name"],"type":"object"}}]}'
|
||||
headers:
|
||||
User-Agent:
|
||||
- X-USER-AGENT-XXX
|
||||
accept:
|
||||
- application/json
|
||||
accept-encoding:
|
||||
- ACCEPT-ENCODING-XXX
|
||||
anthropic-version:
|
||||
- '2023-06-01'
|
||||
connection:
|
||||
- keep-alive
|
||||
content-length:
|
||||
- '1960'
|
||||
content-type:
|
||||
- application/json
|
||||
host:
|
||||
- api.anthropic.com
|
||||
x-api-key:
|
||||
- X-API-KEY-XXX
|
||||
x-stainless-arch:
|
||||
- X-STAINLESS-ARCH-XXX
|
||||
x-stainless-async:
|
||||
- 'false'
|
||||
x-stainless-lang:
|
||||
- python
|
||||
x-stainless-os:
|
||||
- X-STAINLESS-OS-XXX
|
||||
x-stainless-package-version:
|
||||
- 0.71.1
|
||||
x-stainless-retry-count:
|
||||
- '0'
|
||||
x-stainless-runtime:
|
||||
- CPython
|
||||
x-stainless-runtime-version:
|
||||
- 3.12.10
|
||||
x-stainless-timeout:
|
||||
- NOT_GIVEN
|
||||
method: POST
|
||||
uri: https://api.anthropic.com/v1/messages
|
||||
response:
|
||||
body:
|
||||
string: '{"model":"claude-sonnet-4-20250514","id":"msg_01CjWBqSxyLeArjUhqUTbedN","type":"message","role":"assistant","content":[{"type":"text","text":"The
|
||||
document is a minimal PDF file with basic structure containing one empty page
|
||||
with standard letter dimensions (612x792 points).\n\nJVBERi0xLjQKMSAwIG9iaiA8PCAvVHlwZSAvQ2F0YWxvZyAvUGFnZXMgMiAwIFIgPj4gZW5kb2JqCjIgMCBvYmogPDwgL1R5cGUgL1BhZ2VzIC9LaWRzIFszIDAgUl0gL0NvdW50IDEgPj4gZW5kb2JqCjMgMCBvYmogPDwgL1R5cGUgL1BhZ2UgL1BhcmVudCAyIDAgUiAvTWVkaWFCb3ggWzAgMCA2MTIgNzkyXSA+PiBlbmRvYmoKeHJlZgowIDQKMDAwMDAwMDAwMCA2NTUzNSBmCjAwMDAwMDAwMDkgMDAwMDAgbgowMDAwMDAwMDU4IDAwMDAwIG4KMDAwMDAwMDExNSAwMDAwMCBuCnRyYWlsZXIgPDwgL1NpemUgNCAvUm9vdCAxIDAgUiA+PgpzdGFydHhyZWYKMTk2CiUlRU9GCg=="}],"stop_reason":"end_turn","stop_sequence":null,"usage":{"input_tokens":1035,"cache_creation_input_tokens":0,"cache_read_input_tokens":0,"cache_creation":{"ephemeral_5m_input_tokens":0,"ephemeral_1h_input_tokens":0},"output_tokens":400,"service_tier":"standard"}}'
|
||||
headers:
|
||||
CF-RAY:
|
||||
- CF-RAY-XXX
|
||||
Connection:
|
||||
- keep-alive
|
||||
Content-Type:
|
||||
- application/json
|
||||
Date:
|
||||
- Fri, 23 Jan 2026 19:08:58 GMT
|
||||
Server:
|
||||
- cloudflare
|
||||
Transfer-Encoding:
|
||||
- chunked
|
||||
X-Robots-Tag:
|
||||
- none
|
||||
anthropic-organization-id:
|
||||
- ANTHROPIC-ORGANIZATION-ID-XXX
|
||||
anthropic-ratelimit-input-tokens-limit:
|
||||
- ANTHROPIC-RATELIMIT-INPUT-TOKENS-LIMIT-XXX
|
||||
anthropic-ratelimit-input-tokens-remaining:
|
||||
- ANTHROPIC-RATELIMIT-INPUT-TOKENS-REMAINING-XXX
|
||||
anthropic-ratelimit-input-tokens-reset:
|
||||
- ANTHROPIC-RATELIMIT-INPUT-TOKENS-RESET-XXX
|
||||
anthropic-ratelimit-output-tokens-limit:
|
||||
- ANTHROPIC-RATELIMIT-OUTPUT-TOKENS-LIMIT-XXX
|
||||
anthropic-ratelimit-output-tokens-remaining:
|
||||
- ANTHROPIC-RATELIMIT-OUTPUT-TOKENS-REMAINING-XXX
|
||||
anthropic-ratelimit-output-tokens-reset:
|
||||
- ANTHROPIC-RATELIMIT-OUTPUT-TOKENS-RESET-XXX
|
||||
anthropic-ratelimit-tokens-limit:
|
||||
- ANTHROPIC-RATELIMIT-TOKENS-LIMIT-XXX
|
||||
anthropic-ratelimit-tokens-remaining:
|
||||
- ANTHROPIC-RATELIMIT-TOKENS-REMAINING-XXX
|
||||
anthropic-ratelimit-tokens-reset:
|
||||
- ANTHROPIC-RATELIMIT-TOKENS-RESET-XXX
|
||||
cf-cache-status:
|
||||
- DYNAMIC
|
||||
request-id:
|
||||
- REQUEST-ID-XXX
|
||||
strict-transport-security:
|
||||
- STS-XXX
|
||||
x-envoy-upstream-service-time:
|
||||
- '5453'
|
||||
- '100630'
|
||||
status:
|
||||
code: 200
|
||||
message: OK
|
||||
|
||||
Reference in New Issue
Block a user