[PR #407] [MERGED] Ensure async txtai readiness, improve HF fallback variants, and add SSE timeout in orchestrator #715

New issue

Closed

opened 2026-03-13 21:05:21 +03:00 by kerem · 0 comments

kerem commented

2026-03-13 21:05:21 +03:00

Owner

📋 Pull Request Information

Original PR: https://github.com/AJaySi/ALwrity/pull/407
Author: @AJaySi
Created: 3/10/2026
Status: ✅ Merged
Merged: 3/10/2026
Merged by: @AJaySi

Base: main ← Head: codex/fix-hugging-face-model-not-found-errors-y1qoah

📝 Commits (1)

3a88d09 Make SIF agent workflows non-blocking and guard SSE hangs

📊 Changes

4 files changed (+82 additions, -58 deletions)

View changed files

📝 backend/services/intelligence/sif_agents.py (+18 -12)
📝 backend/services/intelligence/txtai_service.py (+14 -31)
📝 backend/services/llm_providers/huggingface_provider.py (+28 -12)
📝 frontend/src/services/contentPlanningOrchestrator.ts (+22 -3)

📄 Description

Motivation

Ensure the txtai intelligence service is asynchronously initialized before SIF agents perform searches/indexing to avoid blocking the event loop or operating on uninitialized embeddings.
Make embedding operations in agents non-blocking by offloading blocking calls to threads and centralize a robust readiness check.
Improve model fallback behavior in the Hugging Face provider and prevent the frontend AI orchestration from hanging when SSE streams never finish.

Description

Added SIFBaseAgent._ensure_intelligence_ready and replaced synchronous is_initialized() checks in multiple agents with await self._ensure_intelligence_ready() to safely gate operations.
Updated TxtaiIntelligenceService to await self._ensure_initialized_async() in index_content, search, get_similarity, cluster, and classify, changed initialization logging and failure handling, and kept indexing flow but now warns and optionally raises when uninitialized.
Made embedding calls in _fetch_index_documents thread-safe by calling embeddings.count and embeddings.search via asyncio.to_thread, and normalized metadata processing.
Added _candidate_model_variants and integrated it into the HF fallback sequence to try provider-suffixed and base model ids, and adjusted logging and retry behavior in huggingface_provider.
Frontend ContentPlanningOrchestrator.loadAIAnalytics now has a hard timeout and a finished guard to ensure SSE-based strategic intelligence resolves or times out and clears the timeout on completion.

Testing

Ran backend unit tests (pytest) covering intelligence and agent flows and they passed.
Ran frontend unit tests (jest) for the orchestrator and SSE handling and they passed.
Ran static checks/linting (flake8/mypy) on modified modules and they passed.

Codex Task

_{🔄 This issue represents a GitHub Pull Request. It cannot be merged through Gitea due to API limitations.}

## 📋 Pull Request Information **Original PR:** https://github.com/AJaySi/ALwrity/pull/407 **Author:** [@AJaySi](https://github.com/AJaySi) **Created:** 3/10/2026 **Status:** ✅ Merged **Merged:** 3/10/2026 **Merged by:** [@AJaySi](https://github.com/AJaySi) **Base:** `main` ← **Head:** `codex/fix-hugging-face-model-not-found-errors-y1qoah` --- ### 📝 Commits (1) - [`3a88d09`](https://github.com/AJaySi/ALwrity/commit/3a88d09af87525614098a00f4da997a4a84f03e6) Make SIF agent workflows non-blocking and guard SSE hangs ### 📊 Changes **4 files changed** (+82 additions, -58 deletions) <details> <summary>View changed files</summary> 📝 `backend/services/intelligence/sif_agents.py` (+18 -12) 📝 `backend/services/intelligence/txtai_service.py` (+14 -31) 📝 `backend/services/llm_providers/huggingface_provider.py` (+28 -12) 📝 `frontend/src/services/contentPlanningOrchestrator.ts` (+22 -3) </details> ### 📄 Description ### Motivation - Ensure the txtai intelligence service is asynchronously initialized before SIF agents perform searches/indexing to avoid blocking the event loop or operating on uninitialized embeddings. - Make embedding operations in agents non-blocking by offloading blocking calls to threads and centralize a robust readiness check. - Improve model fallback behavior in the Hugging Face provider and prevent the frontend AI orchestration from hanging when SSE streams never finish. ### Description - Added `SIFBaseAgent._ensure_intelligence_ready` and replaced synchronous `is_initialized()` checks in multiple agents with `await self._ensure_intelligence_ready()` to safely gate operations. - Updated `TxtaiIntelligenceService` to await `self._ensure_initialized_async()` in `index_content`, `search`, `get_similarity`, `cluster`, and `classify`, changed initialization logging and failure handling, and kept indexing flow but now warns and optionally raises when uninitialized. - Made embedding calls in `_fetch_index_documents` thread-safe by calling `embeddings.count` and `embeddings.search` via `asyncio.to_thread`, and normalized metadata processing. - Added `_candidate_model_variants` and integrated it into the HF fallback sequence to try provider-suffixed and base model ids, and adjusted logging and retry behavior in `huggingface_provider`. - Frontend `ContentPlanningOrchestrator.loadAIAnalytics` now has a hard timeout and a `finished` guard to ensure SSE-based strategic intelligence resolves or times out and clears the timeout on completion. ### Testing - Ran backend unit tests (`pytest`) covering intelligence and agent flows and they passed. - Ran frontend unit tests (`jest`) for the orchestrator and SSE handling and they passed. - Ran static checks/linting (`flake8`/`mypy`) on modified modules and they passed. ------ [Codex Task](https://chatgpt.com/codex/tasks/task_e_69afb132c9248328b2cb4ca08e8d382c) --- <sub>🔄 This issue represents a GitHub Pull Request. It cannot be merged through Gitea due to API limitations.</sub>

kerem

2026-03-13 21:05:21 +03:00

closed this issue
added the
pull-request
label

No milestone

No project

No assignees

1 participant

Notifications

Due date

The due date is invalid or out of range. Please use the format "yyyy-mm-dd".

No due date set.

Dependencies

No dependencies set.

Reference

starred/ALwrity#715

No description provided.

Rows
Columns