Preserves per-thread conversation history and enforces model call budgets.
Java 17+, Maven 3.9.1+, and the matching local ARGI/Extensions RC2 artifacts. Set ARGI_API_KEY and ARGI_MODEL; optionally set ARGI_BASE_URL to an OpenAI-compatible /v1 endpoint.
See the repository setup before running a module.
From the repository root, after installing the example reactor:
bash tools/run.sh argi-agent-example/context-engineeringTwo turns share customer-42. A third model turn in that thread exceeds the configured budget; a different thread has its own budget.
- ContextEngineering.java is the primary entry point.
- Website coverage maps documentation chapters to examples.
- Compatibility notes describe RC2 API changes and the verification boundary.
Local builds and selected runtime checks were performed on 2026-10-15. See validation results for the exact tested scope and remaining failures; compilation alone does not certify an external integration.
- Preserves per-thread conversation history and enforces model call budgets.
- Interleaved customer threads demonstrate isolation and per-thread budget exhaustion.
Limit: In-memory checkpoints and model-call limits; no cross-process session durability.
| Source | Responsibility |
|---|---|
| ContextEngineering.java | Maintains per-thread history and limits the total number of model turns. |
| ConversationSessions.java | Reuses one agent while keeping each customer's checkpoint history separate. |
Run from the repository root after installing the reactor dependencies. Select an entry point explicitly. The validation record distinguishes completed runs from unexecuted or failed scenarios.
Reuses one agent while keeping each customer's checkpoint history separate. Runtime: model.
bash tools/run.sh argi-agent-example/context-engineering -Dexample.mainClass=io.github.agentic.ai.examples.ConversationSessions