Terraform state recovery, plus the version reporting fix.

State recovery

The workspace holding terraform.tfstate was lost. With no backend block the state was local only, so a plan or apply from a fresh checkout would have started from empty state and tried to create the namespace, both databases, the OpenSearch cluster and the app that are already running.

- New terraform/bootstrap-state.sh rebuilds state by importing what is already deployed. Import only, never applies, so it cannot create, replace or destroy anything.
- It regenerates terraform.tfvars from the values in the running pods. This matters because epo_consumer_key, epo_consumer_secret and mistral_api_key default to an empty string in variables.tf, so applying without them would blank out live credentials and break EPO and AI search.
- State has been rebuilt for all ten kubernetes and helm resources. Verified after: PVCs still bound, app pod still running with zero restarts, and a targeted plan reports 0 to add, 8 to change, 0 to destroy. No replacements, so no volume was at risk.

Version reporting

generate-version.js initialised version to a hardcoded constant while both fallback branches still tested for a different sentinel, so the git tag and package.json lookups were dead code and every build reported that constant. That is why the app reported 0.3.2 while the newest release was v0.3.0 and package.json said 0.3.1. Release 0.4.0 was itself stamped 0.3.2 for the same reason. This is the first release that reports its own version correctly.

Still needs a human

- grafana_auth is not discoverable from the bpperegister namespace, so the grafana resources are not yet imported.
- terraform/locals.tf pins the app image at sha256:f22e4589 while the cluster is running sha256:b0a14979. Applying as is would change the running image. Reconcile that digest before any apply.

Document consultation and improvements

This release adds document consultation so you can open BPP patent documents (e.g. PDFs) directly from the app, plus links to classification documentation and improvements to patent results and infrastructure.

---

Document consultation

View patent documents from the detail page
• When you open a BPP (national patent) result, the Available documents section now lists the attached documents (claims, descriptions, etc.) with clickable links.
• Clicking a link opens the document in a new tab. Documents are stored in a private bucket; the app generates short-lived secure links (presigned URLs) so you can view them without exposing storage credentials.
• If document access is not configured in your environment, the document list still appears but links are not shown — no change to how you browse patent metadata.

---

Classification and reference links

Easier to look up codes
• IPC and CPC documentation: From patent results and search, you can open the official IPC (WIPO International Patent Classification) and CPC (Cooperative Patent Classification, EPO) documentation to understand or look up classification codes.
• These links are available in the patent detail view and in the unified search interface so you can quickly check what a code means.

---

Patent results and dependencies

Clearer display and upkeep
• Patent results layout: The patent detail view has been refined so information and document lists are easier to scan and use.
• Dependencies: Project dependencies have been updated for security and stability.
• Backend and AI: Small improvements to the BPP service and to the search assistant’s behaviour for more consistent results.

---

Infrastructure

• Terraform: OpenSearch ingress rules and Terraform configuration (including database and monitoring) have been updated for deployment.
• Documentation: Internal documentation has been updated to reflect recent changes.

---

Summary

• Document consultation: Open BPP patent documents (PDFs, etc.) from the patent detail page via secure, short-lived links.
• Classification docs: Quick links to official IPC (WIPO) and CPC (EPO) documentation from the UI.
• Polish: Better patent results display, dependency updates, and infrastructure/ops updates.

Thank you for using eRegister.

Search, pagination, and layout improvements

This release improves search reliability and pagination, fixes layout and prerendering issues, and adds tracing for better observability.

---

Search and pagination

Faster and more reliable results
• Page caching: Search results are now cached per page so pagination is faster and more consistent when you move between pages.
• Pagination fixes: Page numbers and result ranges now display correctly in both simple and advanced search.
• LLM search: Natural-language (AI) search now paginates correctly so you can browse all results.
• BPP filtering: Improved filtering of BPP (national patent) data so results match your query better.
• IPC search: Fixed search by IPC (International Patent Classification) so classification-based queries return the right results.
• Document details after NLP search: Opening a document from natural-language search now shows the correct patent details.

---

Layout and stability

Smoother experience
• Hydration fix: The layout no longer flickers or misaligns when the page loads; server and client renders now match.
• Build fix: Version information loads correctly during static build, resolving prerendering errors in Server Components.

---

Observability

• Tracing: Tracing has been added to help monitor and debug the application in production.

---

Summary

• Search: Page caching, pagination fixes (including for AI search), better BPP filtering, IPC search fix, and correct document details after NLP search.
• Stability: Hydration and static prerendering fixes for the Next.js layout.
• Operations: Tracing enabled for better observability.

Thank you for using eRegister.

BPP data via OpenSearch and simple indexation

This release connects eRegister to the BPP internal patent register using OpenSearch and a simple indexation model so you can search Belgian, Dutch, and Luxembourg national patents (and related rights) directly from the app.

---

Connection to BPP data using OpenSearch

Unified search over BPP data
• The app now talks to an OpenSearch cluster that hosts the BPP patent index. When OpenSearch is configured, simple and advanced searches for national patents (BE, NL, LU), national patent types (NP), SPC/UP, and related rights are served from this index instead of external APIs.
• Simple search (number, title, applicant, inventor, right type, dates) and advanced search (query builder with multiple conditions) both run as OpenSearch queries against the patents index, with proper mapping of filters (e.g. jurisdiction, right type, filing/publication/grant/lapse dates) and pagination.

Routing and fallback
• The search API automatically routes requests: queries that only involve EPO-specific types (EPV, PCT, US) still use the EPO OPS API; all other queries use BPP (OpenSearch) when it is available. If OpenSearch is not configured, the app falls back to EPO-only behaviour.
• The AI/MCP search assistant uses the same BPP index when OpenSearch is available, so natural-language and conversational searches also benefit from the internal register data.

Configuration
• Connection is controlled by environment variables: OPENSEARCH_URL, optional OPENSEARCH_INDEX (default patents), OPENSEARCH_USERNAME / OPENSEARCH_PASSWORD, and OPENSEARCH_TLS_REJECT_UNAUTHORIZED for TLS. No code changes are needed to switch between environments.

---

Simple indexation of the data

Single index, clear structure
• BPP data is stored in a single OpenSearch index (by default patents) with a straightforward document shape: publication/application numbers, title, applicants, inventors, agent, jurisdiction, right type, IPC/main IPC, status, key dates (filing, publication, grant, expiration, lapse, priority), procedural language, abstract/claims/description text, and attached documents. Semicolon-separated lists (e.g. applicants, IPC codes) are normalized for display in the UI.

Operational workflow
• Indexation is kept simple: data is prepared and indexed into OpenSearch (or restored from snapshots). The app does not perform the indexation itself; it only reads from the index. A restore-from-snapshot workflow (e.g. from S3 to a Kubernetes OpenSearch cluster) is supported via the terraform/restore-on-k8s.sh script, so the same index can be restored in different environments for consistent search behaviour.

---

Small UI updates

Interface polish
• Layout and navigation: Small tweaks to spacing, alignment, and navigation so the app feels a bit cleaner and easier to scan.
• Labels and tooltips: Clearer labels on search options and action buttons; helpful tooltips added where they explain what a control does.
• Consistency: Minor visual and interaction fixes across search and results so behaviour is more consistent.

---

Summary

• OpenSearch integration: BPP patent search (simple + advanced) and AI/MCP search use the OpenSearch patents index when configured.
• Simple indexation: One index, one document type, read-only consumption and optional snapshot restore for deployment.
• Small UI updates: Layout polish, clearer labels and tooltips, and consistency improvements across the interface.

Thank you for using eRegister.

Release 0.2.3

This release improves how the assistant understands and answers your questions, adds feedback and analytics features, and polishes search and the interface.

---

Smarter answers from the assistant

Improved “brain” of the search assistant
We refactored the part of the system that turns your questions into the right searches and answers. You don’t need to do anything different — the assistant should now interpret your requests more reliably and give clearer, more consistent responses. Think of it as a tune-up under the hood: same experience, better behaviour.

---

What’s better for you

Feedback and insights
• Satisfaction survey: You can share how satisfied you are with search results, so we can keep improving the tool.
• Analytics: We collect anonymous usage data to understand how the app is used and where to focus improvements.

Search and results
• Search suggestions: The app can suggest related or alternative searches to help you find what you need.
• Better classification search: Searches by IPC and CPC (patent classification codes) work more reliably.
• More accurate date search: Queries that involve dates (filing, publication) return more accurate results.

Interface and stability
• Simpler layout: The layout has been simplified and fixed so the interface is easier to use.
• Banner and tooltips: The disclaimer banner is clearer, and breaking/action buttons have helpful tooltips.
• Clean-up: Unused features (e.g. RSS feed) have been removed; dependencies have been updated for security and stability.

Thank you for using eRegister. We’re continuing to improve the assistant and search based on your feedback.

Release 0.2.2

This release makes it easier to find the patents you care about and improves how the tool understands your questions.

---

What’s better for you

New ways to search
• Search by application number: You can now look up a patent directly using its application number — handy when you have a reference from another document or database.
• US and PCT patents: The app now properly supports US patents and international (PCT) documents across search and display, so you get consistent results when working with different patent types.

Smarter search and filters
• Better date filters: Publication and filing date filters work more reliably, so when you search by date (e.g. “filed after 2020” or “published in 2024”) you get the results you expect.
• Improved natural language: The tool understands your questions better and turns them into the right search, with fewer odd or unexpected results.
• Conversational search: When you use the chat-style search, named searches (e.g. by inventor or applicant) work more smoothly and the experience feels more stable.

More reliable results
• Total count and scrolling: The total number of results is shown correctly, and scrolling through long result lists (including infinite scroll) behaves as expected.
• Default search region: When searching in BENELUX or Europe, the default region is applied correctly so you don’t have to adjust settings unnecessarily.
• Clearer warnings: If you run an EPO search by legal status, you’ll see a clear warning so you know what to expect.

Small fixes that add up
• Country names: Countries with special characters or accents are displayed and searched correctly.
• Country and region handling: The app converts country names to the right codes behind the scenes, so your searches are more accurate.

Thank you for using eRegister. We’re continuing to improve search and will keep you updated.

Release 0.2.1

This release continues to improve search reliability and your experience when looking up patents.

Our focus going forward
The search backend has been improved, but our main priority is to get better at understanding your questions and turning them into the right query. We are also preparing to make the most of BPP data as it becomes available.

---

What’s better for you

Search and results
• More reliable searches: Searches behave more consistently and date filters (e.g. publication or filing date) are applied correctly more often.
• Fewer search errors: The system handles your query in a more robust way, so you see fewer unexpected errors.
• Pagination and feedback: Small fixes so page numbers and feedback work as expected.

When you search by inventor or applicant
• Name variations: If nothing is found at first, the tool can try other name formats (e.g. last name only, or "Montfort") so you get results when they exist in our data.
• Clearer “no results”: You get suggestions you can try (e.g. “Try: Montfort”) and, for US or other out-of-scope patents, a short explanation and links to USPTO or Google Patents.

Interface
• Release notes: You can see what’s new in each version on the release notes page.
• Favicon: The app now has its own icon in the browser tab.

Thank you for using eRegister. We’re working on improving how the tool understands your questions and will keep you updated.

AI Search Improvements (18 user reports addressed)

Search reliability
• Smarter search: The tool now correctly understands when you search by inventor, applicant, patent number, or date — and searches the right fields instead of only the title.
• Name variations: For inventor and applicant names, the system tries different name formats automatically (e.g. "Montfort" or variations) so you get results even when names are stored differently.
• When nothing is found: You now get clear suggestions (e.g. try "Montfort" only) and, for US or other out-of-scope patents, a short explanation and links to USPTO or Google Patents.

Better understanding of your question
• Dates: Phrases like "patents filed after 2010" or "filed in 2020" are now correctly interpreted as filing date filters.
• Companies and applicants: Queries like "Schlumberger" or "applicant X" are treated as applicant/company searches, not mixed with title search.

Data and interface
• Document types: WO (PCT) documents are now labelled as "International (PCT)" instead of "National patent".
• Links: The "Document" and "Espacenet" links now open the correct patent document.
• Pagination: Page numbers at the bottom of the results list now display correctly.

Thank you for your feedback — these changes are based on your reports.