Deterministic + LLM-powered Production Log Diagnosis for JVM / Spring / Kafka
A deterministic Java log analyzer for Spring Boot, Hibernate and Kafka, powered by local LLMs (Ollama) for safe production diagnostics.
Log Doctor is a production-grade log analysis tool that detects failures in JVM-based systems
(Spring Boot, Hibernate, Kafka) and provides precise root-cause analysis and safe fixes.
It combines:
- ⚙️ Deterministic rule-based detection (HIGH confidence incidents)
- 🧠 Local LLM reasoning via Ollama (for unknown or ambiguous logs)
🧩 Designed for real production systems, not toy examples.
Modern JVM applications produce massive logs, but:
- Stacktraces are noisy
- Root causes are buried
- Most tools only display logs — they don’t understand them
Log Doctor answers two critical questions:
- 🔍 Where exactly is the error? (component + layer)
- 🧯 What is the safest possible fix? (only if one exists)
❌ No generic advice
❌ No theory
❌ No hallucinated fixes
Known production failures are detected with HIGH confidence, without LLM guessing:
- Hibernate LazyInitializationException
- Spring NoSuchBeanDefinitionException
- Spring profile mismatches
- Jackson / JSON deserialization failures
- Kafka topic not found
- Kafka schema incompatibility
- HikariCP timeouts
- Deadlocks & thread starvation
- OutOfMemoryError
- GC thrashing
📄 Full list: 👉 docs/supported-errors.md
Each incident includes:
- Category
- Severity
- Confidence
- Allowed Fix Types (policy-driven)
Each supported error is implemented as an independent deterministic rule
under the rules/ package and can be enabled, disabled or extended without
affecting the rest of the system.
For unknown failures, Log Doctor uses a local LLM via Ollama to:
- Identify the deepest application-level root cause
- Decide whether a safe automatic fix exists
- Refuse to propose fixes when human investigation is required
Every fix is constrained by a FixPolicy:
| Category | Allowed Fixes |
|---|---|
| DATABASE | JAVA_CODE |
| CONFIGURATION | SPRING_CONFIG |
| DESERIALIZATION | JAVA_CODE / SPRING_CONFIG |
| INFRASTRUCTURE | KAFKA_CLI / NO_AUTOMATIC_FIX |
| THREADING | JAVA_CODE / NO_AUTOMATIC_FIX |
The LLM cannot violate these rules.
Raw Logs
↓
LogParser
↓
FailureLocator (root cause + blame location)
↓
IncidentDetector (deterministic rules)
↓
┌───────────────┬─────────────────────┐
│ Known Incident│ Unknown Failure │
│ (HIGH CONF) │ │
│ LLM Prompt │ LLM Prompt │
│ (constrained) │ (safe reasoning) │
└───────────────┴─────────────────────┘
- Logs are parsed line by line
- The deepest application-level failure is selected
- Deterministic rules are evaluated first
- LLM is used only if no rule matches
Raw log → Root cause → Blame location → Fix (or refusal)
Input Log
2026-05-10 10:14:33.412 ERROR [http-nio-8080-exec-4] o.s.web.servlet.DispatcherServlet :
Servlet.service() for servlet [dispatcherServlet] in context with path [] threw exception
org.springframework.web.util.NestedServletException: Request processing failed
at org.springframework.web.servlet.FrameworkServlet.processRequest(FrameworkServlet.java:1014)
at org.springframework.web.servlet.FrameworkServlet.doGet(FrameworkServlet.java:898)
Caused by: org.hibernate.LazyInitializationException:
failed to lazily initialize a collection of role:
com.mycompany.myservice.domain.User.orders, could not initialize proxy - no Session
at org.hibernate.collection.internal.AbstractPersistentCollection.throwLazyInitializationException(AbstractPersistentCollection.java:614)
at org.hibernate.collection.internal.AbstractPersistentCollection.withTemporarySessionIfNeeded(AbstractPersistentCollection.java:218)
at org.hibernate.collection.internal.AbstractPersistentCollection.initialize(AbstractPersistentCollection.java:591)
at org.hibernate.collection.internal.AbstractPersistentCollection.read(AbstractPersistentCollection.java:149)
at com.mycompany.myservice.service.UserService.toDto(UserService.java:74)
at com.mycompany.myservice.api.UserController.getUser(UserController.java:52)
at java.base/jdk.internal.reflect.DirectMethodHandleAccessor.invoke(DirectMethodHandleAccessor.java:104)
Output
WHERE:
- component: UserService
- layer: SERVICE
- method or line: UserService.toDto(UserService.java:74)
FIX_TYPE: JAVA_CODE
FIX:
@Transactional
public UserDto toDto(User user) {
return new UserDto(user.getId(), user.getName(),
Optional.ofNullable(user.getOrders()).orElse(Collections.emptyList()));
}
Some failures must not be auto-fixed, e.g.:
- Optimistic locking conflicts
- Concurrent updates
- Cross-transaction data consistency issues
No safe automatic fix – human investigation required.
This is a feature, not a limitation.
ollama pull llama3Other supported models:
- mistral
- codellama
- llama3:instruct
ollama serveOllama runs on:
http://localhost:11434
java -jar log-doctor-0.1.0.jar --file examples/app.loglog-doctor/
├── core/
├── engine/
├── llm/
├── rules/
└── examples/
- Backend engineers (Java / Spring)
- Platform & DevOps engineers
- Kafka & microservices teams
- Anyone debugging production logs at 3AM
- Determinism before AI
- Safety before automation
- Local-first, privacy-first
- Production realism over demos
Apache 2.0 License — use it, extend it, improve it.
log doctor, spring boot logs, java stacktrace analyzer, kafka deserialization error,
hibernate lazy initialization, ollama local llm, deterministic log diagnosis,
spring bean missing, java production debugging, gc thrashing analysis
If your tool:
- Always proposes a fix → ❌ It’s lying
- Never refuses →
⚠️ It’s dangerous - Explains theory only → 💤 It’s not production-ready
🩺 Log Doctor does none of the above.