Observed Signal · Jun 14, 2026 · Technical Release · Source: DEV Community · Impact: 2/5 · Sentiment: Positive
Structured JSON Output from Local LLMs with Ollama & Zod
A technical guide explaining how to produce reliable, validated JSON from small local LLMs using Ollama and Zod. The article shows two Ollama JSON modes — format: "json" for syntactic validity and passing a full JSON Schema to format to constrain structure — and recommends converting Zod schemas to JSON Schema with zod-to-json-schema so one schema serves both generation and runtime validation. It documents strategies for handling truncated output (raise num_predict, repair by closing open brackets), streaming (accumulate newline-delimited envelopes and parse when complete), and retry loops that feed Zod validation errors back into the model while lowering temperature. The author provides a reusable generateStructured<T>(schema) helper combining schema-constrained generation, repair, and converging retries, and notes this pattern is used in spectr-ai for local smart-contract audits.
Provides practical, reproducible techniques for making local LLM outputs reliably structured and validated — useful for privacy-sensitive or on-device workflows but not a major platform policy change.
Track Ollama Signals & Market Shifts in Real-Time
Polaris7 autonomous intelligence agents track regulatory filings, primary sources, executive changes, and deal flow 24/7. Create your free Explorer workspace to monitor these entities.
Key Takeaways & Evidence Grounding
- Ollama supports format: "json" to enforce syntactically valid JSON output from models.
- Ollama also accepts a full JSON Schema as format to constrain generated output to a structure.
- Zod schemas can be converted to JSON Schema via zod-to-json-schema to keep one source of truth.
- Truncated JSON can sometimes be repaired by trimming to the last complete token and closing open brackets; increasing num_predict (e.g., 2048) reduces truncation.
- Retry loops that feed specific Zod validation errors back to the model and adjust temperature help small local models converge to valid output.
Connected Companies & Entities
1 Entity mappedOntology Mapping & Concepts
Related Market Signals & Shifts
Recent verified developments and strategic activity across this market segment.
Structured Output: Treat Schema as a Contract
INTFRAME published a technical blog post arguing that LLM outputs consumed by machines should be treated as enforceable contracts: model responses must conform to JSON schemas validated by ordinary validators. Their production loop retries invalid outputs (feeding validator errors back verbatim) up to three times, with items failing validation moved to a human-reviewed quarantine. Key tactics include using enums instead of free text, setting temperature to 0, and applying constrained decoding where supported to reduce hallucinations and increase reliability.
SmarterJSON: Reader for Messy JSON and LLM Output
A developer argues that traditional JSON parsers are overly strict and discard usable data when input deviates by a single byte (trailing commas, BOMs, comments, etc.). The article documents recurring real-world failure modes — NDJSON, LLM-generated "almost-JSON", duplicate keys, and high-precision numbers — and contrasts recognition (strict parsing) with extraction (robust data recovery). The author presents SmarterJSON, an open-source JSON processor (github.com/tilo/smarter_json) designed to read a superset of JSON in one pass, preserve high-precision numbers, return typed data, report any fixes, and avoid inventing missing data. The post calls for readers whose default is lenient extraction rather than strict grammar recognition to reduce production incidents caused by malformed or dialect-variant JSON.
Build Zero-Crash LLM JSON Pipelines Without Regex
The article presents a production-grade approach to avoid fragile regex-based JSON extraction from LLM outputs. It argues that most pipeline failures come from malformed JSON (trailing commas, truncated strings, unescaped quotes) and proposes a Three-Layer Validation Pattern: pre-sanitization, strict schema binding (using Pydantic), and a targeted repair fallback that re-routes malformed output to a fast repair model. The author provides example code using OpenAI's structured outputs with a Pydantic model and gives operational advice: check the API's finish_reason, avoid manual regex for parsing, and use cheap sub-second models to repair truncated or invalid JSON responses.
Track Real-Time Market Signals & Shifts
Set up custom watchlists to receive automated, evidence-grounded executive digests whenever material signals or shifts occur across your tracked landscape.