JSON
A lightweight data-interchange format that uses curly braces to wrap key-value pairs and square brackets to denote arrays. Its clear structure makes it easy for both machines to parse and humans to read, and it has become the de facto standard for Web API communication.
Core Facts
The basic principle of tool calling is that the model identifies scenarios requiring external capabilities, outputs a structured function call request (usually in JSON format), and after the external system executes it, the result is returned to the model to continue reasoning.
Function Call does not actually execute code; the model only outputs a structured JSON string, and the actual execution is performed by the host program.
Protocol Buffers reduces size by approximately 60–80% and parses 3–10 times faster compared to JSON.
Timeline (last 90 days)
The core flow of the tool is: scan PDF, PNG, and JPG files in a local directory, parse the receipts with a locally deployed open source vision/text model, extract vendor name, date, amount, and expense category, and output structured JSON, all without calling cloud APIs and without needing an internet connection.
By serializing the current state locally before each step (such as to JSON or SQLite) and recovering from the most recent snapshot on failure, one can obtain fault tolerance close to that of a workflow engine without introducing a complex retry framework.
The choice of structured data output format should match the downstream use: JSON is more suitable for consumption by front-end applications, while CSV or Excel is more direct for data analysts to use.
In a test with regular tabular data of 10 rows and 4 fields per row, encoded using the cl100k_base tokenizer, JSON was 323 tokens and 963 characters, while YAML was 221 tokens and 590 characters, with YAML ahead by 31.6%; the TOON format was 97 tokens and 235 characters, 70% fewer than JSON and 56.1% fewer than YAML.
In a test with a nested configuration object, encoded using OpenAI's cl100k_base tokenizer, JSON was 99 tokens and 322 characters, while YAML was 66 tokens and 223 characters, with YAML reducing the token count by 33.3%.
A practical engineering pattern is: keep the data actually flowing inside the system as JSON, convert only the text handed to the model into YAML or TOON, and then require the model to return in JSON format anything that needs to go back into code.
Any data that is to land in an API, a database, or existing software should continue to use JSON, and conversion should occur only at the boundary of interaction with the model.
Python 机器人通常通过 WebSocket 或 TCP 将每个策略的净值、持仓、成交事件序列化为 JSON 流式发送，Unity 端解析后驱动对应建筑对象的属性。
The OCR workflow includes steps such as starting from a clear image, extracting relevant information, converting to a suitable format (such as JSON, CSV, XML, YAML), validating the extraction results, and connecting to applications.
The rewritten prompt returned a completely consistent JSON structure across five runs, allowing results to be fed directly into n8n nodes without any additional repair steps.
37 more timeline events
All Facts (20)
The basic principle of tool calling is that the model identifies scenarios requiring external capabilities, outputs a structured function call request (usually in JSON format), and after the external system executes it, the result is returned to the model to continue reasoning.
75%VerifiedFunction Call does not actually execute code; the model only outputs a structured JSON string, and the actual execution is performed by the host program.
75%VerifiedProtocol Buffers reduces size by approximately 60–80% and parses 3–10 times faster compared to JSON.
75%VerifiedJSON organizes information into key-value pairs that flow through n8n workflow connections.
70%VerifiedEach extracted test point uses a unified JSON format containing the module chapter, test point title, test status and operation steps, expected results, priority, and test design method.
70%VerifiedSwitching to the Input tab in the n8n node editing panel allows you to preview in real time the complete JSON structure flowing into that node
65%VerifiedThe underlying implementation of tool calling typically relies on the Function Calling protocol, where developers describe a tool's name, parameter types, and semantics in JSON Schema format, and the model outputs structured invocation instructions conforming to the protocol format.
65%Verifiedn8n's underlying architecture is built on Node.js, workflow data is stored in JSON format, and the execution engine supports both serial and parallel task scheduling.
65%VerifiedEach Skill declares its own functional description, API endpoint, authentication method, and parameter specifications through a configuration file (typically in YAML or JSON format).
65%VerifiedWeChat Cloud Database stores data in JSON format, with no need to predefine table structures
65%Unverified四层分区架构中，第二层结构化档案使用结构化JSON对象作为单一事实来源（SSOT），采用状态覆写机制处理数据更新
95%UnverifiedJSON规范（RFC 8259）不允许尾随逗号（trailing comma），这与JavaScript对象不同
95%UnverifiedCompared to JSON, protobuf has a smaller encoding size and faster serialization/deserialization speed.
70%UnverifiedJSON相比XML或Protocol Buffers，在LLM提示工程中的可读性和token效率都更优
70%UnverifiedThe core flow of the tool is: scan PDF, PNG, and JPG files in a local directory, parse the receipts with a locally deployed open source vision/text model, extract vendor name, date, amount, and expense category, and output structured JSON, all without calling cloud APIs and without needing an internet connection.
50%UnverifiedBy serializing the current state locally before each step (such as to JSON or SQLite) and recovering from the most recent snapshot on failure, one can obtain fault tolerance close to that of a workflow engine without introducing a complex retry framework.
50%UnverifiedThe choice of structured data output format should match the downstream use: JSON is more suitable for consumption by front-end applications, while CSV or Excel is more direct for data analysts to use.
50%UnverifiedIn a test with regular tabular data of 10 rows and 4 fields per row, encoded using the cl100k_base tokenizer, JSON was 323 tokens and 963 characters, while YAML was 221 tokens and 590 characters, with YAML ahead by 31.6%; the TOON format was 97 tokens and 235 characters, 70% fewer than JSON and 56.1% fewer than YAML.
50%UnverifiedIn a test with a nested configuration object, encoded using OpenAI's cl100k_base tokenizer, JSON was 99 tokens and 322 characters, while YAML was 66 tokens and 223 characters, with YAML reducing the token count by 33.3%.
50%UnverifiedAny data that is to land in an API, a database, or existing software should continue to use JSON, and conversion should occur only at the boundary of interaction with the model.
50%