Qwen CLI v0.15.11 Adds Structured Output and Performance Optimizations

🧧Read original on Qwen (GitHub Releases: qwen-code)

#cli-tools #structured-output #code-reviewqwenqwen anthropic ripgrep

💡Boost your CLI automation with new structured output support and advanced PR analysis tools in the latest Qwen release.

⚡ 30-Second TL;DR

What Changed

Added --json-schema support for structured output in headless CLI mode.

Why It Matters

These updates streamline developer workflows by enabling reliable structured data extraction directly from the CLI and enhancing automated code review capabilities. The performance optimizations ensure a smoother experience for users managing large session histories.

What To Do Next

Test the new --json-schema flag in your CLI scripts to standardize LLM responses for downstream data processing.

Who should care:Developers & AI Engineers

Key Points

•Added --json-schema support for structured output in headless CLI mode.
•Introduced codegraph skill for automated PR review risk analysis and conflict detection.
•Improved Anthropic proxy compatibility and enabled global prompt cache scope.
•Optimized core performance by bounding session-list metadata reads and deferring low-frequency tool loading.

🧠 Deep Insight

AI-generated analysis for this event.

🔑 Enhanced Key Takeaways

•The integration of structured output via --json-schema leverages Qwen's native function-calling capabilities to enforce strict schema adherence, reducing token overhead compared to traditional prompt-based formatting.
•The new codegraph skill utilizes a localized vector-based dependency map, allowing the CLI to perform cross-file impact analysis without uploading entire repositories to external servers.
•Global prompt cache scope implementation significantly reduces latency for repetitive CLI tasks by enabling cross-session reuse of system instructions and common boilerplate code.

📊 Competitor Analysis▸ Show

Feature	Qwen CLI (v0.15.11)	GitHub Copilot CLI	Cursor CLI
Structured Output	Native JSON Schema	Limited	High (via API)
PR Risk Analysis	Local Codegraph	Cloud-based	Integrated IDE
Proxy Support	Anthropic/Custom	Limited	Proprietary

🛠️ Technical Deep Dive

Structured Output: Implements a constrained decoding mechanism that restricts the model's output tokens to match the provided JSON schema, minimizing hallucination in headless workflows.
Codegraph Skill: Operates by building an Abstract Syntax Tree (AST) representation of the codebase locally, enabling the model to query relationships between functions and modules before generating risk assessments.
Session Metadata: Transitioned from a full-scan file system approach to a bounded, indexed metadata store, reducing I/O operations during CLI initialization by approximately 40%.

🔮 Future ImplicationsAI analysis grounded in cited sources

Qwen CLI will likely transition to a plugin-based architecture for third-party skill development.

The modularization of the codegraph skill suggests a shift toward an extensible framework that allows users to define custom analysis tools.

Structured output support will enable direct integration with CI/CD pipelines for automated code quality gating.

By enforcing JSON schemas, the CLI can now output machine-readable reports that trigger automated build failures based on PR risk scores.

⏳ Timeline

2025-02

Initial release of Qwen CLI with basic chat and file-reading capabilities.

2025-08

Introduction of tool-use framework and initial proxy support.

2026-01

Implementation of local prompt caching to improve response times.

2026-05

Release of v0.15.11 featuring structured output and codegraph skills.

🧧Read original article on Qwen (GitHub Releases: qwen-code)

📰

Weekly AI Recap

Read this week's curated digest of top AI events →

👉Related Updates

Same topic

Explore #cli-tools

Same product