[PR #5236] [MERGED] Intelligent Skill Selection #5365

Closed
opened 2026-06-05 15:21:07 -04:00 by yindo · 0 comments
Owner

📋 Pull Request Information

Original PR: https://github.com/Mintplex-Labs/anything-llm/pull/5236
Author: @timothycarambat
Created: 3/19/2026
Status: Merged
Merged: 3/19/2026
Merged by: @timothycarambat

Base: masterHead: tool-reranker


📝 Commits (5)

📊 Changes

37 files changed (+989 additions, -52 deletions)

View changed files

📝 docker/.env.example (+12 -1)
📝 frontend/src/locales/ar/common.js (+19 -0)
📝 frontend/src/locales/cs/common.js (+19 -0)
📝 frontend/src/locales/da/common.js (+19 -0)
📝 frontend/src/locales/de/common.js (+19 -0)
📝 frontend/src/locales/en/common.js (+19 -0)
📝 frontend/src/locales/es/common.js (+19 -0)
📝 frontend/src/locales/et/common.js (+19 -0)
📝 frontend/src/locales/fa/common.js (+19 -0)
📝 frontend/src/locales/fr/common.js (+19 -0)
📝 frontend/src/locales/he/common.js (+19 -0)
📝 frontend/src/locales/it/common.js (+19 -0)
📝 frontend/src/locales/ja/common.js (+19 -0)
📝 frontend/src/locales/ko/common.js (+19 -0)
📝 frontend/src/locales/lv/common.js (+19 -0)
📝 frontend/src/locales/nl/common.js (+19 -0)
📝 frontend/src/locales/pl/common.js (+19 -0)
📝 frontend/src/locales/pt_BR/common.js (+19 -0)
📝 frontend/src/locales/ro/common.js (+20 -0)
📝 frontend/src/locales/ru/common.js (+19 -0)

...and 17 more files

📄 Description

Pull Request Type

  • feat (New feature)
  • 🐛 fix (Bug fix)
  • ♻️ refactor (Code refactoring without changing behavior)
  • 💄 style (UI style changes)
  • 🔨 chore (Build, CI, maintenance)
  • 📝 docs (Documentation updates)

Description

Intelligent Skill Selection is a unique AnythingLLM feature that can save you up to 80% per request when using agent tooling. Right now in the industry every tool is injected into every request. This is extremely wasteful for cloud and problematic for local - where context preservation is key.

ISS utilities AnythingLLM's unique infra and pipeline to assemble and rerank your tools to a topN that from benchmarks achieves a 98% accuracy for even complex tool call chains.

For local model users, this allows you to preserve context for actual chats while keeping prefill and TTFT low and having "unlimited" tools on hand.

For cloud, this means you are saving money every chat.

All of this, for a slight latency benefit (depends on hardware and total tool count). Eg: on M4 Pro we can rerank 300 tools to top 15 in ~600ms. A small price to pay IMO.

Feature is opt-in and will only trigger if the # of active tools > topN for the ISS setting.

Visuals (if applicable)

Adds a menu on Agent skills to:

  • Enable Intelligent Skill Selection
  • Set max tool availability (def: 15)
  • Set max tool call chain limit (def: 10)
settings-menu

Additional Information

Docs: https://github.com/Mintplex-Labs/anythingllm-docs/pull/221

Developer Validations

  • I ran yarn lint from the root of the repo & committed changes
  • Relevant documentation has been updated (if applicable)
  • I have tested my code functionality
  • Docker build succeeds locally

🔄 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/Mintplex-Labs/anything-llm/pull/5236 **Author:** [@timothycarambat](https://github.com/timothycarambat) **Created:** 3/19/2026 **Status:** ✅ Merged **Merged:** 3/19/2026 **Merged by:** [@timothycarambat](https://github.com/timothycarambat) **Base:** `master` ← **Head:** `tool-reranker` --- ### 📝 Commits (5) - [`c72c31b`](https://github.com/Mintplex-Labs/anything-llm/commit/c72c31bb9b8b793bf55a8ca2513d7dbe2c4c51dc) Beta Intelligent Tooling - [`308fb97`](https://github.com/Mintplex-Labs/anything-llm/commit/308fb9721d8253cc58fba124b4096b52041b7333) forgot files - [`be09998`](https://github.com/Mintplex-Labs/anything-llm/commit/be09998c41bd8331829eaaf1badf9409a5f81ac4) add UI controls and maxToolCallStack setting - [`c0576c9`](https://github.com/Mintplex-Labs/anything-llm/commit/c0576c9a9d68f07a91516055f465ba01c63f32d6) update docs link - [`5d8073d`](https://github.com/Mintplex-Labs/anything-llm/commit/5d8073ddf1c6f714cd256d55c8be5233080ce441) ISS i18n (#5237) ### 📊 Changes **37 files changed** (+989 additions, -52 deletions) <details> <summary>View changed files</summary> 📝 `docker/.env.example` (+12 -1) 📝 `frontend/src/locales/ar/common.js` (+19 -0) 📝 `frontend/src/locales/cs/common.js` (+19 -0) 📝 `frontend/src/locales/da/common.js` (+19 -0) 📝 `frontend/src/locales/de/common.js` (+19 -0) 📝 `frontend/src/locales/en/common.js` (+19 -0) 📝 `frontend/src/locales/es/common.js` (+19 -0) 📝 `frontend/src/locales/et/common.js` (+19 -0) 📝 `frontend/src/locales/fa/common.js` (+19 -0) 📝 `frontend/src/locales/fr/common.js` (+19 -0) 📝 `frontend/src/locales/he/common.js` (+19 -0) 📝 `frontend/src/locales/it/common.js` (+19 -0) 📝 `frontend/src/locales/ja/common.js` (+19 -0) 📝 `frontend/src/locales/ko/common.js` (+19 -0) 📝 `frontend/src/locales/lv/common.js` (+19 -0) 📝 `frontend/src/locales/nl/common.js` (+19 -0) 📝 `frontend/src/locales/pl/common.js` (+19 -0) 📝 `frontend/src/locales/pt_BR/common.js` (+19 -0) 📝 `frontend/src/locales/ro/common.js` (+20 -0) 📝 `frontend/src/locales/ru/common.js` (+19 -0) _...and 17 more files_ </details> ### 📄 Description ### Pull Request Type <!-- For change type, change [ ] to [x]. --> - [x] ✨ feat (New feature) - [ ] 🐛 fix (Bug fix) - [x] ♻️ refactor (Code refactoring without changing behavior) - [x] 💄 style (UI style changes) - [ ] 🔨 chore (Build, CI, maintenance) - [x] 📝 docs (Documentation updates) ### Description Intelligent Skill Selection is a unique AnythingLLM feature that can save you up to 80% **per request** when using agent tooling. Right now in the industry every tool is injected into every request. This is extremely wasteful for cloud and problematic for local - where context preservation is key. ISS utilities AnythingLLM's unique infra and pipeline to assemble and rerank your tools to a `topN` that from benchmarks achieves a 98% accuracy for even complex tool call chains. For local model users, this allows you to preserve context for actual chats while keeping prefill and TTFT low and having "unlimited" tools on hand. For cloud, this means you are saving money every chat. All of this, for a slight latency benefit (depends on hardware and total tool count). Eg: on M4 Pro we can rerank 300 tools to top 15 in ~600ms. A small price to pay IMO. Feature is _opt-in_ and will only trigger if the # of active tools > topN for the ISS setting. <!-- Describe the changes in this PR that are impactful to the repo. What problem does it solve? --> ### Visuals (if applicable) Adds a menu on Agent skills to: - Enable Intelligent Skill Selection - Set max tool availability (def: 15) - Set max tool call chain limit (def: 10) <img width="1718" height="895" alt="settings-menu" src="https://github.com/user-attachments/assets/866a7ce7-71ab-4ef1-acb2-359c53d810b7" /> <!-- Add screenshots or screen recordings to demonstrate the changes, especially for UI updates. --> ### Additional Information Docs: https://github.com/Mintplex-Labs/anythingllm-docs/pull/221 <!-- Add any other context about the Pull Request here that was not captured above. --> ### Developer Validations <!-- All of the applicable items should be checked. --> - [ ] I ran `yarn lint` from the root of the repo & committed changes - [ ] Relevant documentation has been updated (if applicable) - [ ] I have tested my code functionality - [ ] Docker build succeeds locally --- <sub>🔄 This issue represents a GitHub Pull Request. It cannot be merged through Gitea due to API limitations.</sub>
yindo added the pull-request label 2026-06-05 15:21:07 -04:00
yindo closed this issue 2026-06-05 15:21:07 -04:00
Sign in to join this conversation.
1 Participants
Notifications
Due Date
No due date set.
Dependencies

No dependencies set.

Reference: Mintplex-Labs/anything-llm#5365