v1.1.0:2 — model-agnostic AI program generation (5 providers)

Five providers behind one streaming abstraction: - claude (Anthropic) - openai (api.openai.com) - openai-compatible (any base URL — OpenRouter / LiteLLM / vLLM / Together / your own gateway) - gemini (Google) - ollama (self-hosted; no key; LAN URL like http://ollama.embassy:11434) The "self-hosted Ollama on Start9" angle is the killer use case — configure Settings → AI integration with the LAN URL of your Ollama service and no API keys ever leave your network. Architecture - lib/ai/types.ts LLMProvider streaming interface - lib/ai/sse.ts shared SSE + NDJSON line iterators - lib/ai/providers/*.ts 5 implementations + factory - lib/ai/programSchema.ts Zod schema + JSON-schema-for-prompt + parseAIProgram with markdown-fence stripping and balanced-brace JSON extraction - lib/ai/apply.ts materializes parsed AIProgram into Program tree (validates exerciseIds, rejects unresolved nulls, atomic transaction, sets aiGenerated=true) Schema - UserPreferences gets aiProvider/aiModel/aiBaseUrl/aiApiKey (plaintext — same threat model as the rest of /data). Dead enableClaudeAI/claudeApiKey columns from v1.0.0:1-7 stay as no-op fields. - AIPromptTemplate (userId nullable; userId=NULL = built-in) - AIGeneration (raw response + parsed program + status + appliedProgramId + token counts) - All compat-ALTER'd in docker_entrypoint.sh on first boot. API - POST /api/ai/generate SSE streaming: emits generation/text/usage/complete events; persists AIGeneration row up front so failures show in history too - POST /api/ai/apply takes user-edited AIProgram, creates Program, marks generation as applied - GET /api/ai/templates built-ins + this user's own - POST /api/ai/templates create user-owned template - PATCH /api/ai/templates/[id] edit; built-ins admin-only - DELETE /api/ai/templates/[id] delete; built-ins admin-only - GET /api/ai/generations list (paginated) - GET /api/ai/generations/[id] full row - DELETE /api/ai/generations/[id] delete one (Program survives) - GET /api/ai/config returns aiKeyConfigured flag, never plaintext key - POST /api/ai/config update provider config - DELETE /api/admin/ai/generations admin-only "clear all" with optional userId / olderThanDays UI - Settings → AI integration provider/model/URL/key form; plaintext key warning visible - /main/ai hub page with cards - /main/ai/generate template picker + textarea + live SSE stream + cancel + ProgramPreview with inline unknown-exercise resolver + apply button + redirect to the new Program - /main/ai/templates list + create + edit + delete; per-row "show prompt" expand; built-in delete warns about reconcile re-creation - /main/ai/history list + delete; status badges; link to applied Program - Nav: "AI" entry between Programs and Exercises (Sparkles icon) Built-in templates - prisma/aiTemplates.seed.json: 5 starter templates (hypertrophy / strength / endurance / recovery / custom) - prisma/ensurePromptTemplates.cjs: per-boot reconcile, INSERT-or-UPDATE keyed on (userId IS NULL AND name=...); user-created templates never touched Tests - tests/ai-programSchema.test.ts: extractJson + parseAIProgram edge cases (markdown fences, balanced braces, malformed JSON, Zod shape rejection, unresolved-exerciseId tolerance) - tests/ai-apply.test.ts: materializes valid AIProgram, rejects cross-user exerciseIds, rejects unresolved exercises, honors isActive flag - tests/routes-ai-templates.test.ts: built-in vs user permissions, cross-user template isolation, /api/ai/config plaintext-key safety, provider enum validation - 123 tests across 14 files, all passing. No data migration. Existing /data is augmented with the new columns + tables only.
2026-05-10 15:35:35 -05:00
parent 3a5b929284
commit 974c3eb07d
36 changed files with 4206 additions and 1 deletions
@@ -7,6 +7,7 @@ import { v_1_0_0_5 } from './v1.0.0.5'
 import { v_1_0_0_6 } from './v1.0.0.6'
 import { v_1_0_0_7 } from './v1.0.0.7'
 import { v_1_1_0_1 } from './v1.1.0.1'
+import { v_1_1_0_2 } from './v1.1.0.2'

 /**
 * Version graph for the `proof-of-work` package.
@@ -22,9 +23,11 @@ import { v_1_1_0_1 } from './v1.1.0.1'
 * v1.0.0:6 — paginate workout history (infinite scroll).
 * v1.0.0:7 — exercise library cleanup, photo-import removal.
 * v1.1.0:1 — Programs UI (manual create / save / follow).
+ * v1.1.0:2 — AI program generation, 5 providers (Claude / OpenAI /
+ *            OpenAI-compatible / Gemini / Ollama).
 */
 export const versionGraph = VersionGraph.of({
-  current: v_1_1_0_1,
+  current: v_1_1_0_2,
  other: [
    v_1_0_0_1,
    v_1_0_0_2,
@@ -33,5 +36,6 @@ export const versionGraph = VersionGraph.of({
    v_1_0_0_5,
    v_1_0_0_6,
    v_1_0_0_7,
+    v_1_1_0_1,
  ],
 })
@@ -0,0 +1,62 @@
+import { IMPOSSIBLE, VersionInfo } from '@start9labs/start-sdk'
+
+/**
+ * v1.1.0:2 — model-agnostic AI program generation.
+ *
+ * Five providers behind one streaming abstraction:
+ *
+ *   - claude              (Anthropic)
+ *   - openai              (api.openai.com)
+ *   - openai-compatible   (any base URL — OpenRouter / LiteLLM /
+ *                          vLLM / Together / your own gateway)
+ *   - gemini              (Google)
+ *   - ollama              (self-hosted; no key; uses LAN URL)
+ *
+ * The "self-hosted Ollama on Start9" angle is the killer use case —
+ * point Settings → AI integration → Base URL at
+ * `http://ollama.embassy:11434` (or whatever Ollama service you have
+ * on the same StartOS host) and no API keys ever leave your network.
+ *
+ * Workflow
+ *   1. Settings → AI integration: pick provider + model + key/URL.
+ *   2. AI → Generate program: pick a template, type your specifics
+ *      ("8 weeks heavy leg emphasis"), click Generate.
+ *   3. Watch the response stream in word-by-word via SSE.
+ *   4. Server validates the JSON output against a Zod schema and
+ *      stores both raw + parsed in AIGeneration.
+ *   5. Preview UI shows the program tree. Unknown exercises (the
+ *      model picked something not in your library) are highlighted
+ *      and you can map them to existing entries or remove them.
+ *   6. Apply → materializes into a real Program (same schema/UI as
+ *      the v1.1.0:1 manual programs).
+ *
+ * Ships with 5 built-in prompt templates (hypertrophy block,
+ * strength block, endurance/running block, recovery week, custom).
+ * Built-ins reconcile per-boot the same way curated exercises do.
+ * Both admin and regular users can create their own templates.
+ *
+ * Schema additions
+ *   - UserPreferences: aiProvider, aiModel, aiBaseUrl, aiApiKey
+ *     (plaintext — consistent with the rest of /data/app.db; the
+ *     host-level threat model assumes the operator owns /data).
+ *   - AIPromptTemplate (built-ins userId=NULL, user templates
+ *     userId=<them>).
+ *   - AIGeneration (one row per generate request; raw response,
+ *     parsed program, status, applied program id, token counts).
+ *
+ * Backward compatible: existing UserPreferences rows get the new
+ * columns added with NULL defaults (compat ALTER on first boot);
+ * the dead enableClaudeAI / claudeApiKey columns from v1.0.0:1-7
+ * stay as no-op fields.
+ */
+export const v_1_1_0_2 = VersionInfo.of({
+  version: '1.1.0:2',
+  releaseNotes: {
+    en_US:
+      'AI program generation. Pick from Claude / OpenAI / Gemini / OpenAI-compatible / self-hosted Ollama. Settings → AI integration to configure (Ollama on Start9 needs no API key). AI → Generate program to pick a template, describe what you want, watch the response stream in, review the parsed program, and apply it to your Programs library. Ships 5 starter templates; both admin and regular users can create their own. Generation history is kept until you delete it (per-row Trash; admin-only "clear all" via /api/admin/ai/generations).',
+  },
+  migrations: {
+    up: async () => {},
+    down: IMPOSSIBLE,
+  },
+})