More dependable context. Easier adoption.
Better evidence where engineering decisions happen.
This is our proposed direction, drawn from the product backlog. Themes are grouped by readiness, not delivery order. Scope and priorities may change; planned and exploratory capabilities are not delivery commitments.
Make everyday answers more dependable and the path to a working installation more direct.
Show when a search covers only part of the corpus, improve representation across repositories, and clarify how disabled repositories affect results.
Preserve work-item provider identity and useful links when individual source records cannot be read.
Extend incremental graph processing across indexing paths and make sustained drift and repeated indexing failures visible at repository level.
Build on the hosted-Git graph improvements already in the codebase, with clearer configuration and recovery signals.
Publish the CLI package, simplify access to release images, make compiled reports easier to find, and help AI clients recognise when to use Engram.
Public image distribution still needs a commercial decision. A guided demo remains the current starting point.
Make Helm installations more adaptable to existing secret managers, workload identities, and optional Redis or Ollama services.
Add a reusable post-upgrade check that verifies retrieval as well as service health.
Open feature work with dependencies, validation, or rollout decisions still to resolve.
Extend the GitHub annotation foundation to Azure DevOps, GitLab, and Bitbucket, with an optional informational GitHub check.
Each integration remains separately opt-in. Strengthen delivery recovery and make annotation coverage clearer.
Bring Contract Sentinel evidence into PR review so teams can inspect a changed API surface against its indexed baseline.
Extend the existing post-index capability. Report what extraction supports, with explicit limits when evidence is incomplete.
Add previous-release compatibility checks, data-preserving Compose upgrade tests, and an executable upgrade and rollback guide.
Optional update notices and customer release access remain dependent on distribution and operating decisions.
Develop local usage attribution, report-only seat and renewal views, and exportable evidence of repository access and indexing triggers.
Respect repository scope and human versus service identities. Audit exports support a review process; they do not imply certification.
Demand-led ideas that need more evidence or a separate design decision before implementation.
Evaluate complete air-gap bundles and additional release channels when customer network restrictions or rollout needs justify them.
Self-hosting and offline license verification exist today. A complete offline installation bundle is separate future work.
Extend Kotlin and Swift parsing, Python import resolution, and selected PHP and SQL relationships as real codebases demonstrate the need.
These deepen existing language support; extraction coverage varies by language and framework.
Explore turning validated technical-debt findings into bounded draft changes after the issue-filing phase establishes a reliable record.
Separate approval, testing, and rollback controls are prerequisites. Autonomous approval or merging is outside this direction.
Evaluate partitioning data, jobs, credentials, and caches before considering a shared multi-tenant deployment.
Today's repository access controls apply within a managed corpus. They are not a shared SaaS isolation model.
Bring a real codebase question to a demo. It helps us understand which improvements matter in practice.