Tessera v1.5.0 — Moat Phase

Moat Phase — 5 strategic directions for competitive defensibility

What's new

• NL Time Query — bilingual (EN+KO) natural language time expressions: "yesterday", "지난주", "3월 초" auto-detected in recall and search
• Context Priming — session-start briefing with recent decisions, preferences, active topics, last session summary
• Provenance Chain — track origin, session, and parent lineage of every memory
• Project Spaces — isolate knowledge by project scope, recall filtered by project
• Quiet Curation — background maintenance on MCP server startup: classify, tag, extract entities, resolve contradictions, consolidate, flag retention
• Auto-Insights — trending topics, decision patterns, hidden connections, weekly digest
• Product Landing Page — tessera-dev.vercel.app

Numbers

Metric	Count
MCP tools	58
HTTP endpoints	54
CLI commands	11
Core modules	69
Lines of code	31,000+
Tests	1,102 (all passing)
File types	40+

Links

• Website: https://tessera-dev.vercel.app
• PyPI: https://pypi.org/project/project-tessera/
• Full changelog: https://github.com/besslframework-stack/project-tessera/blob/main/CHANGELOG.md

Tessera v1.1.0 — Insight Phase

Insight Phase — From passive storage to active knowledge intelligence

Tessera now actively analyzes your memories, detects contradictions, and searches from multiple angles. Cross-project patterns adopted from Claudel, Fleming, Tether, and channeltalk-mcp.

New MCP Tools & HTTP Endpoints (+6 each)

Feature	MCP Tool	HTTP Endpoint
Multi-angle Search	deep_search	POST /deep-search
Multi-angle Recall	deep_recall	POST /deep-recall
Contradiction Detection	detect_contradictions	GET /contradictions
Memory Confidence	memory_confidence	GET /memory-confidence
Memory Health	memory_health	GET /memory-health
Plugin Hooks	list_plugin_hooks	GET /hooks

Search Verdict

Every search/recall result now includes a confidence label: confident match (≥45%), possible match (25-45%), low relevance (<25%).

By the numbers

• 53 MCP tools (+6)
• 34 HTTP endpoints (+6)
• 784 tests (+132)
• 6 new modules

Install / upgrade

pip install --upgrade project-tessera

v1.0.1 — Production/Stable

Fixed

• Tool count correction — MCP tool count was incorrectly stated as 51; corrected to 47 (@mcp.resource is not a tool)
• README and CHANGELOG updated with accurate numbers

Changed

• PyPI classifier upgraded from Beta to Production/Stable

Tessera v1.0.0 — Personal Knowledge Layer for AI

Tessera v1.0 GA

Tessera is a personal knowledge layer that works across every AI tool.

What it does:

• Remembers decisions, preferences, and facts across AI sessions
• Works with Claude (MCP), ChatGPT, Gemini, and any HTTP-capable AI tool
• Exports to Obsidian, Markdown, CSV, JSON, ChatGPT memory, Gemini context
• Imports past conversations from ChatGPT, Claude, and Gemini
• Encrypts memories at rest with AES-256 (optional)
• 100% local: LanceDB + fastembed/ONNX, no cloud, no API keys

By the numbers:

• 51 MCP tools
• 28 HTTP API endpoints
• 11 CLI commands
• 652 tests
• 4 phases: Sponge → Radar → Gateway → Cortex

Install:

pip install project-tessera

Full changelog: CHANGELOG.md

v0.9.0 — Gateway Phase Complete

Gateway Phase complete — Tessera works with any AI tool via REST API.

What's new (v0.8.1 → v0.9.0)

• FastAPI HTTP API with 21 endpoints
• API key authentication (optional, via TESSERA_API_KEY)
• CORS for ChatGPT, Gemini, Claude.ai origins
• Batch API (up to 20 operations in one call)
• Rate limiting (configurable, default 60 req/min)
• Webhooks for event-driven integrations

Stats

• 44 MCP tools + 21 HTTP endpoints
• 483 tests passing
• 40+ supported file types
• Python 3.11+

Install: pip install project-tessera==0.9.0
HTTP API: pip install project-tessera[api]==0.9.0 && tessera-api

v0.8.0 — Radar Phase Complete

Radar Phase complete — Tessera now has proactive intelligence.

What's new (v0.7.1 → v0.8.0)

• Universal text/code parser (50+ file extensions)
• Temporal index with date/category filters on recall
• Decision tracker — groups decisions by topic, detects changes
• Context window builder — optimal context assembly within token budget
• Smart suggest — personalized query recommendations
• Relevance decay — time-based score adjustment for memories
• Topic map — keyword clustering with Mermaid mindmap
• Knowledge stats — aggregate statistics dashboard

Stats

• 44 MCP tools
• 413 tests passing
• 40+ supported file types

Install: pip install project-tessera==0.8.0

v0.6.4 — Interaction Log

Sponge Phase 1/7: Automatic logging of every MCP tool call.

Added

• Interaction Log — every tool call recorded to SQLite (tool, input, output, duration, session ID)
• session_interactions tool — view current/past session activity
• recent_sessions tool — session history with interaction counts
• Total tools: 33, tests: 193

Install

pip install project-tessera==0.6.4

v0.6.3

Added

• TESSERA_WORKSPACE env var — set document folder path via environment variable, enables uvx users to specify workspace without workspace.yaml
• 3 new config tests (total: 184)

Install

pip install project-tessera

Or with uvx:

uvx --from project-tessera tessera-mcp

v0.6.0

First PyPI release

pip install project-tessera

Added

• PyPI package — install with pip install project-tessera
• tessera setup — one-command setup (workspace.yaml + model download + Claude Desktop config)
• Auto-detect workspace — works without workspace.yaml
• PDF support — optional dep (pip install project-tessera[pdf])
• Search highlighting — matching terms bold-wrapped in results
• Friendly error messages — action-oriented, no jargon
• tessera check improvements — index size, model cache, config validation
• 31 tools | 181 tests

v0.5.2

Added

• PDF file support — pymupdf-based parser, optional dep (pip install -e ".[pdf]")
• Search result highlighting — matching terms bold-wrapped in search results
• tessera check improvements — LanceDB index size, model cache, Claude Desktop config validation, dependency checks
• Total tools: 31 | Tests: 181