feat: implement AtoCore Phase 0 + Phase 0.5 (foundation + PoC)

Complete implementation of the personal context engine foundation:
- FastAPI server with 5 endpoints (ingest, query, context/build, health, debug)
- SQLite database with 5 tables (documents, chunks, memories, projects, interactions)
- Heading-aware markdown chunker (800 char max, recursive splitting)
- Multilingual embeddings via sentence-transformers (EN/FR)
- ChromaDB vector store with cosine similarity retrieval
- Context builder with project boosting, dedup, and budget enforcement
- CLI scripts for batch ingestion and test prompt evaluation
- 19 unit tests passing, 79% coverage
- Validated on 482 real project files (8383 chunks, 0 errors)

Co-Authored-By: Claude Opus 4.6 (1M context) <noreply@anthropic.com>

tests/test_prompts/gigabit_prompts.yaml

@@ -0,0 +1,40 @@
+prompts:
+  - id: g1
+    prompt: "What is the GigaBIT M1 project about?"
+    project: gigabit
+    expected: "Should mention 1.2m primary mirror, StarSpec, telescope"
+
+  - id: g2
+    prompt: "What are the main requirements for the M1 mirror?"
+    project: gigabit
+    expected: "Should mention optical/mechanical requirements, SOW, diameter, Zerodur"
+
+  - id: g3
+    prompt: "What vendors are involved in the project?"
+    project: gigabit
+    expected: "Should mention Optiques Fullum, StarSpec, Atomaste, or subcontractors"
+
+  - id: g4
+    prompt: "What is the status of the CDR?"
+    project: gigabit
+    expected: "Should mention Critical Design Review status, CBUSH, design completion"
+
+  - id: g5
+    prompt: "What are the key design decisions made so far?"
+    project: gigabit
+    expected: "Should mention design phases, PDR, assumptions, blank order"
+
+  - id: g6
+    prompt: "What FEA optimization work has been done?"
+    project: gigabit
+    expected: "Should mention FEA analysis, optimization approach, WFE, displacement data"
+
+  - id: g7
+    prompt: "What is the cost reduction strategy?"
+    project: gigabit
+    expected: "Should mention cost reduction campaign, trade-off, topology selection"
+
+  - id: g8
+    prompt: "What are the mirror blank specifications?"
+    project: gigabit
+    expected: "Should mention 1200mm diameter, Zerodur, optical specifications"

tests/test_prompts/prompts.yaml

@@ -0,0 +1,40 @@
+prompts:
+  - id: p1
+    prompt: "What is AtoCore's architecture?"
+    project: atocore
+    expected: "Should mention layered architecture, SQLite, vector DB"
+
+  - id: p2
+    prompt: "What chunking strategy does AtoCore use?"
+    project: atocore
+    expected: "Should mention heading-aware splitting, 800 char max"
+
+  - id: p3
+    prompt: "What is the trust precedence order?"
+    project: atocore
+    expected: "Should list: Trusted Project State > AtoDrive > validated memory"
+
+  - id: p4
+    prompt: "How does AtoCore handle conflicts between sources?"
+    project: atocore
+    expected: "Should mention conflict resolution rules, no silent merging"
+
+  - id: p5
+    prompt: "What are the different memory types?"
+    project: atocore
+    expected: "Should list: Identity, Preferences, Project, Episodic, Knowledge, Adaptation, Trusted Project State"
+
+  - id: p6
+    prompt: "What is the context budget allocation?"
+    project: atocore
+    expected: "Should mention percentages: identity 5%, preferences 5%, project 20%, episodic 10%, retrieval 60%"
+
+  - id: p7
+    prompt: "What is a trivial prompt in AtoCore?"
+    project: atocore
+    expected: "Should mention: no project ref, no proper nouns, no past context dependency"
+
+  - id: p8
+    prompt: "What are the success criteria for the first win?"
+    project: atocore
+    expected: "Should mention: saves >=5 min lookup, >=80-90% accuracy, >=10 test prompts"