v5.5.0

What's Changed

• Add DevSecOps pipeline: SAST, SCA, secret scanning, IaC checks, and Claude review by @gerardrecinto in #305
• feat: Azure Container Apps deployment for tools/httpserver by @gerardrecinto in #306
• test: raise btree/common coverage 85.3% -> 87.7%, fix stale Codecov slug by @gerardrecinto in #307
• release: v5.5.0 by @gerardrecinto in #308
• fix(bindings): bindings/main missing from the Docker build context by @gerardrecinto in #309

Full Changelog: v5.4.0...v5.5.0

v5.4.0 - Human copy pass, security fix, real enterprise contact

Website & Docs

• Rewrote the marketing site copy (demo/index.html) to read like a founder explaining the product, not AI-generated marketing copy: hero, verification barrier, business value, why-now, business model, pricing, enterprise value, enterprise form, final CTA.
• Removed every em dash and leftover buzzword ("paradigm", "mission-critical") from the site copy, and softened an unverifiable "1 business day" enterprise response-time claim.
• README: shrunk the logo from 480px to 120px, added a "Cutting a Release" section documenting the real scripts/update_version.sh / build_release.sh / verify_release.sh flow.

Security

• Bumped google.golang.org/grpc v1.83.1 -> v1.83.2, closing GHSA-2v4p-qf9q-27wj (CVE-2026-84445, xDS server DoS). Transitive dependency only; this repo does not run an xDS gRPC server.

Fixed

• Enterprise contact fallback now points to a real, monitored address instead of a placeholder inbox.

Full Changelog: v5.3.8...v5.4.0

SOP v5.3.8 - Production Resilience, Multi-Language Alignment, and AI Memory Release

Changelog

v5.3.8

Highlights

• Full ecosystem version alignment: Synchronized all core Go modules and language bindings (Python sop4py, C# Sop, Java, Rust) to release version v5.3.8.
• Zero known vulnerabilities: Pinned Go toolchain 1.26.8 across all modules and Docker containers, resolving 7 standard library CVEs and verifying 0 called vulnerabilities with govulncheck.
• Deterministic container and CI test validation: Resolved direct I/O failover simulation in root/container environments and restored full race-detector test coverage in CI.

Reliability

• Container test runner hardening: Updated Test_EC_Failover_Reinstate_FastForward_Short in infs/integrationtests to utilize the DirectIO simulator and fs.TriggerFailover instead of filesystem chmod manipulations, eliminating permission bypasses in container and root execution contexts.
• Race detector coverage: Verified clean execution across core packages under Go race detection (inmemory, btree, common, fs, cache, cel, encoding, database, jsondb, search).

Performance

• Reproducible cache latency: Verified 62.09 ns/op L1 MRU hit latency and 66.17 us/op under heavy mixed-cache workloads.
• Vector store throughput: Benchmarked end-to-end ACID transaction pipeline in ai/vector at 16.23 ms for batch 50-vector upsert, KNN query, and durable commit.
• CI benchmark stabilization: Calibrated benchmark iteration scaling in performance workflows to prevent runner timeouts.

Developer Experience

• Quickstart zero-dependency demo: Validated examples/quickstart and published container image (ghcr.io/sharedcode/sop-quickstart) providing instant zero-configuration testing.
• Unified versioning: Enhanced scripts/update_version.sh for atomic multi-language version updates.

AI/Data

• Knowledge Base compilation: Verified and generated complete base knowledge base in ai/sop_base_knowledge.json containing 1,257 knowledge items and 1,397 categories.
• Query and Join execution: Verified Adaptive Hash Join planning in AI agent runtime executing in 1.10 ms.

Security

• Go toolchain update: Upgraded all modules and Dockerfiles to Go 1.26.8, addressing CVEs in net/url, html/template, crypto/tls, encoding/asn1, net/http, and golang.org/x/net/idna.
• Dependency hardening: Bumped go-git to v5.19.2 and cel-go to v0.29.0 to eliminate open Dependabot advisories.

CI/CD

• Gated delivery pipeline: Decoupled production promotion into .github/workflows/promote.yml so pending manual approvals do not block or fail continuous delivery status.
• Test execution fix: Explicitly configured package test targets in .github/workflows/go.yml for comprehensive coverage reporting with race detector.

Documentation

• Badge cleanup: Removed retired Go Report Card badge, added dynamically resolved Go version, license, and CI build status badges.
• Architecture and Investor analysis: Added detailed architectural positioning ("Why SOP?") and multi-horizon product roadmap.

Breaking Changes

• None. Full backward compatibility maintained for all storage layouts, transaction protocols, and APIs.

Upgrade Notes

• Go 1.26.8 or later is recommended when building from source. No data migration required for existing stores.

v5.3.7

• Descending iterators for the in-memory B-Tree: AllDesc() and RangeDesc(from, to) walk keys newest-first; RangeDesc seeks straight to the high bound. Both covered by unit tests; quickstart shows a newest-first scan.
• Fix: iteration errors no longer swallowed in the Data Manager item stream (tools/httpserver): a shadowed err inside the paging loop meant store.Next failures never reached the error log or terminated the loop condition.
• Fix: tools/confighub knowledge-base tests now skip on fresh clones instead of failing when the gitignored local tools/config.json is absent.
• Security: jackson-databind 2.21.4 to 2.21.5 in the Java binding (closes the case-insensitive deserialization bypass advisory, the last open Dependabot alert).
• Lint cleanup: removed a stray debug print in the item search path, replaced nil Contexts with context.Background() in the in-memory B-Tree wrapper, finished the typed context key for the vectorized-spaces map, normalized error strings, dropped dead nil checks.

v5.3.6

• Maintenance rebuild: fixed v5.3.5/v5.3.6 release pipeline failures by building the whole knowledge_compiler package and bumping the build image to Go 1.26.4. No library code changes.

v5.3.5

• B-Tree node slot allocation and L1 cache handling optimized for better throughput on hot paths.
• L2 cache eviction policy reworked for standalone mode -- smarter eviction under memory pressure.
• NewBtree/OpenBtree safe multi-open: the same named B-Tree can now be opened multiple times within a single transaction without data races. Stress tests updated to cover this.
• Range-over-func iterators for the in-memory B-Tree: All() and Range(from, to) return iter.Seq2 so callers can for k, v := range b3.Range(102, 104). Range seeks straight to the start key; both are covered by unit tests. Quickstart example and demo GIF updated to use them.
• Security: cleared all 32 Dependabot alerts (7 critical). Go: golang.org/x/crypto 0.52.0, golang.org/x/net 0.55.0, go-git/v5 5.19.1, go-billy/v5 5.9.0, cloudflare/circl 1.6.3. Java binding: jackson-databind 2.19.0 to 2.21.4 (fixes CVE PolymorphicTypeValidator bypass, array subtype allowlist bypass, InetSocketAddress SSRF, case-insensitive deserialization bypass).
• Gated delivery pipeline (.github/workflows/deliver.yml): every push to master runs build, tests, container packaging to GHCR (sop-quickstart), and a staging smoke test. Production promotion (image :stable tag plus GitHub Pages site deploy) sits behind a manual approval on the production environment.
• Quickstart example (examples/quickstart): zero-infrastructure in-memory B-Tree walkthrough (add, find, update, ordered scan). Packaged as a distroless container via Dockerfile.quickstart.
• README demo GIF recorded from the quickstart run; project site landing page added (index.md).
• Root directory cleanup: 26 documentation files moved to docs/, 13 shell scripts moved to scripts/. README and workflow references updated. Root now contains only source code, standard project files (LICENSE, CONTRIBUTING, etc.), and build configs.

SOP V2 build 54 (Upcoming)

• Gate 1 Advanced KB Routing: Major enhancements to specialized focused routing for knowledge base queries.
  • Root Category Navigation: Query omni:<kb> to display all root categories with item counts and subcategory information.
    • Example: omni:sop shows all top-level categories in the SOP knowledge base
    • Provides directory-style exploration without needing to know category names upfront
    • Navigation hints included for deeper exploration (e.g., "Navigate: omni:sop:language")
    • Pagination: 20 categories per page with :page:<number> or /page/<number> syntax
      • omni:sop:page:2 or omni:sop/page/2 - View page 2 of root categories
      • omni:sop:language:page:3 or omni:sop/language/page/3 - View page 3 of subcategories
      • Supports both : and / as separators (matches user's query style)
      • Shows page info: "(Page 2 of 5, showing 21-40 of 87)"
      • Navigation hints: "Previous: omni:sop:page:1 | Next: omni:sop:page:3"
      • LLM filtering suggestion for large sets (>40 categories)
  • :llm <instruction> Meta-Token: Added support for explicit LLM post-processing instructions using :llm suffix (e.g., omni:sop:operations:performance:llm summarize top 3).
    • Clean Query Separation: The :llm meta-token is automatically stripped from the KB search query and treated as post-retrieval guidance.
    • TaskContextClassification Fields: Added CleanQuery and LLMInstruction fields to properly separate user intent from meta-commands.
    • Three-Way Routing: Intelligent decision-making based on result count and :llm presence:
      • :llm present → LLM processes with instruction (highest priority)
      • 1-5 matches → Direct display (no LLM)
      • 6+ matches → Automatic LLM summarization
  • Flexible Hierarchy Support: Full support for any-depth category paths (e.g., omni:kb:cat1:subcat1.1:subsubcat1.1.1:...).
  • Subcategory Navigation: When a category path has no direct items (and no :llm instruction), automatically returns child categories with item counts and descriptions as navigation hints.
  • Enhanced Parsing: New stripLLMInstruction() function ensures consistent meta-token extraction across all query patterns.
  • Architecture Improvements:
    • getSubcategories() function for root and path-level category display
    • buildKBEnrichedQuery() now uses clean queries without meta-tokens for proper LLM context
    • trySpecializedFocusedRouting() handles root navigation, flexible hierarchy, and meta-token parsing
    • Comprehensive test coverage for all routing patterns and hierarchy depths
  • Roadmap - Quoted Text Search: Proposed support for combined category + text queries (e.g., omni:sop:language bindings "java tutorial").
  • Documentation Updates: Updated AI_COPILOT.md, AI_COPILOT_USAGE.md, and IMPLEMENTATION.md with comprehensive routing guides including root category navigation.

SOP V2 build 53 (Upcoming)

• Schema Format Enhancement: Introduced flat schema format for better LLM understanding and correlation with Store Relations.
  • New Fields: Added FlatSchema, KeyFields, and ValueFields to StoreInfo for improved schema representation.
  • Flat Schema: Uses flat format without p...

v5.3.7

Full Changelog: v5.3.6...v5.3.7

v5.3.6

What's Changed

• wired: health endpoint for load balancers by @gerardrecinto in #299

Full Changelog: v5.3.5...v5.3.6

SOP v5.3.5

What changed since v5.3.4

Performance

• B-Tree node slot allocation and L1 cache handling got faster. Measured improvement on hot-path operations.
• L2 cache eviction for standalone mode reworked -- smarter about what to keep under memory pressure.
• NewBtree/OpenBtree can now safely be called multiple times for the same B-Tree within a single transaction. Previously this could cause subtle data races. Stress tests updated.

Security -- all Dependabot alerts cleared

• Go deps: golang.org/x/crypto 0.52.0, golang.org/x/net 0.55.0, go-git/v5 5.19.1, go-billy/v5 5.9.0, cloudflare/circl 1.6.3
• Java binding: jackson-databind 2.19.0 -> 2.21.4
  • PolymorphicTypeValidator bypass via generic type parameters (high)
  • Array subtype allowlist bypass in BasicPolymorphicTypeValidator (high)
  • InetSocketAddress deserialization triggers eager DNS resolution / SSRF (medium)
  • Case-insensitive deserialization bypasses per-property @JsonIgnoreProperties (medium)

32 alerts total were open at the start of this cycle. All 32 are now resolved.

In-memory B-Tree iterators

• All() and Range(from, to) now return iter.Seq2 -- you can do for k, v := range b3.Range(102, 104). Range seeks directly to the start key.

CI/CD

• Gated delivery pipeline (.github/workflows/deliver.yml): build -> tests -> GHCR container (sop-quickstart) -> staging smoke -> manual production promotion.
• Quickstart container (Dockerfile.quickstart): distroless image running the in-memory B-Tree demo.

Developer experience

• Quickstart example at examples/quickstart -- no servers, no config, just go run ./examples/quickstart.
• "What is SOP" plain-language overview added and linked from README and site.
• Demo GIF recorded from the quickstart run.

Repository cleanup

• 26 documentation markdown files moved from root to docs/.
• 13 shell scripts moved from root to scripts/.
• All references updated (README, index.md, release workflow, VS Code tasks, DEV_TOOLS.md).
• Root directory now has source code, standard project files, Dockerfiles, and Go module files. That is it.

---

Full changelog: v5.3.4...v5.3.5

SOP V2: general purpose AI & Knowledge Base support PATCH Update

Upgraded to latest go 1.26.4.

SOP V2: general purpose AI & Knowledge Base support PATCH update

Minor cleanup and CICD fix.

SOP V2: general purpose AI & Knowledge Base support

Summary:

• Support for "general purpose" AI like Claude Code, Gemini & Chat GPT
• Support for first class Knowledge Base, a.k.a. Spaces. Mint: visualize, enrich, group your "thoughts" into nested Categories to
• Support million/billions data set without the 'pigeon hole' weakness, then Vectorize them for "semantic" retrieval
• Support for local Embedders, thanks to Kelindar Search Open source folks!
• Support for rich prefix based routing with zero LLM calls, e.g. custom KB:Category1:Category1.1
• Support for LLM filtering, e.g.: e.g. custom KB:Category1:Category1.1:llm [instruction filter]
• Richer Stores schema view and management
• Stores and Spaces can complement each other, write ACID transactions that manage/query the two databases
• Ask LLM to manage the two databases, no need to learn SQL syntax
• Very rich LLM integration implementation taking full advantage of providers' server side continuation/conversation features, and SOP "memory" feature (MRU, Short Term Memory & Long Term Memory structures).
• Cleaner Release Numbering, targeting latest latest so go toolchain will not get confused.
• ai/memory KnowledgeBase package can easily be reused externally. See http://github.com/grecinto/tensortree
• More robust/optimized SOP B-tree engine:
  • Store Repository Count tracking imposes minimal update of store as opposed to entire Store repository metadata overwrite
  • More efficient B-tree storage use, i.e. - it is now practical to use large SlotLengths like 10,000 per Node. It requires same storage space as 500 SlotLength Node
  • Beta 2/Community Preview Stability

SOP V2: general purpose AI & Knowledge Base support patch update

Few minor refactors/updates to allow easier reuse of ai/memory KnowledgeBase package and dependencies.