ui: Context usage gauge and panel by allozaur · Pull Request #25340 · ggml-org/llama.cpp

allozaur · 2026-07-06T06:59:44Z

Overview

demo_h264_2.mp4

Moved Reasoning to Chat Form Add Panel (renamed from ChatFormActionReasoningEffortSubmenu → ChatFormActionAddReasoningSubmenu, nested as a submenu)
Updated defaults for Reasoning (OFF added to ReasoningEffort enum, labels restructured)
Context usage gauge and panel — new donut progress + hover card with model info, context bar, and collapsible token breakdown (tool defs, reading, output, avg speed)
Removed ChatScreenProcessingInfo component and its context provider (replaced by the gauge)
toolsStore now owns the OpenAI wire format exclusively — removed getToolDefinitionsForLLM and normalizeSchemaProperties from mcpStore
/tokenize endpoint + TokenizeService with FNV-1a hashing for tool token count caching
MCP store cleanup — removed inline OpenAI tool conversion, connection lifecycle/health checks/executeTool remain
Stream constants update, migration service updates

Requirements

I have read and agree with the contributing guidelines
AI usage disclosure: YES, Ornith-1.0-35B with Zed, Pi and llama-ui via llama serve

ServeurpersoCom · 2026-07-07T16:58:01Z

This is a genuine optimization of states display and ergonomy. It’s no longer a draft, it's ready for review!
I'm keeping the live branch running on my server to see if I run into any bugs; so far, I have no issues to report for now

Copilot

Pull request overview

Adds a context-usage gauge to the chat UI and refactors related reasoning/tooling plumbing in tools/ui, including reasoning defaults, MCP tool wire-format ownership, and removal of the old processing-info UI.

Changes:

Introduces a new context usage gauge (donut + hover-card details) and supporting hook/components.
Reworks reasoning defaults/controls (adds ReasoningEffort.OFF, updates how thinking is derived and sent to the API).
Refactors MCP tool definition handling so toolsStore owns OpenAI-compatible tool payload shaping; removes the old processing info component/context and some UI settings.

Reviewed changes

Copilot reviewed 52 out of 53 changed files in this pull request and generated 2 comments.

Show a summary per file

File	Description
tools/ui/src/lib/utils/branching.ts	Trims docstrings in branching helpers
tools/ui/src/lib/stores/tools.svelte.ts	Moves MCP tool wire-shaping into tools store
tools/ui/src/lib/stores/mcp.svelte.ts	Removes OpenAI tool generation from MCP store
tools/ui/src/lib/stores/conversations.svelte.ts	Derives thinking from reasoning effort defaults
tools/ui/src/lib/stores/chat.svelte.ts	Omits effort when set to OFF
tools/ui/src/lib/stores/agentic.svelte.ts	Types now reference toolsStore tool payload
tools/ui/src/lib/services/migration.service.ts	Comment tweak in config migration
tools/ui/src/lib/services/chat.service.ts	Minor doc/comment wording change
tools/ui/src/lib/hooks/use-processing-state.svelte.ts	Simplifies monitoring stop behavior
tools/ui/src/lib/hooks/use-mcp-recommendations.svelte.ts	Simplifies opt-in dialog scheduling logic
tools/ui/src/lib/hooks/use-context-gauge.svelte.ts	New hook for context gauge state
tools/ui/src/lib/enums/reasoning-effort.enums.ts	Adds `ReasoningEffort.OFF`
tools/ui/src/lib/enums/index.ts	Re-exports new chat statistics mode enum
tools/ui/src/lib/enums/chat.enums.ts	Adds `ChatMessageStatisticsMode`
tools/ui/src/lib/contexts/processing-info.context.ts	Removes processing-info context (deleted)
tools/ui/src/lib/contexts/index.ts	Stops exporting processing-info context
tools/ui/src/lib/constants/storage.ts	Removes thinking-default localStorage key
tools/ui/src/lib/constants/settings-registry.ts	Removes some display settings entries
tools/ui/src/lib/constants/settings-keys.ts	Removes keys for deleted settings
tools/ui/src/lib/constants/reasoning-effort.ts	Adds OFF label and enum-backed values
tools/ui/src/lib/constants/css-classes.ts	Adds submenu width constant
tools/ui/src/lib/constants/context-keys.ts	Removes processing-info context key
tools/ui/src/lib/components/ui/hover-card/index.ts	New hover-card component exports
tools/ui/src/lib/components/ui/hover-card/hover-card.svelte	New hover-card root wrapper
tools/ui/src/lib/components/ui/hover-card/hover-card-trigger.svelte	New hover-card trigger wrapper
tools/ui/src/lib/components/ui/hover-card/hover-card-portal.svelte	New hover-card portal wrapper
tools/ui/src/lib/components/ui/hover-card/hover-card-content.svelte	New hover-card content wrapper
tools/ui/src/lib/components/app/settings/SettingsChat/SettingsChatToolsTab.svelte	Adjusts tools tab label rendering
tools/ui/src/lib/components/app/dialogs/DialogMcpServerRecommendations.svelte	Updates recommendations dialog behavior
tools/ui/src/lib/components/app/content/SyntaxHighlightedCode.svelte	Improves code block overflow sizing
tools/ui/src/lib/components/app/content/CollapsibleContentBlock.svelte	Adjusts collapsible layout classes
tools/ui/src/lib/components/app/chat/index.ts	Exports new reasoning submenu + gauge; removes old processing info export
tools/ui/src/lib/components/app/chat/ChatScreen/ChatScreenProcessingInfo.svelte	Removes processing info component (deleted)
tools/ui/src/lib/components/app/chat/ChatScreen/ChatScreen.svelte	Drops processing-info usage/context wiring
tools/ui/src/lib/components/app/chat/ChatMessages/ChatMessageStatistics/ChatMessageStatistics.svelte	Adds fixed-mode stats rendering support
tools/ui/src/lib/components/app/chat/ChatMessages/ChatMessages.svelte	Passes last-user/next-assistant metadata
tools/ui/src/lib/components/app/chat/ChatMessages/ChatMessageAgenticContent.svelte	Revises agentic turn layout + stats gating
tools/ui/src/lib/components/app/chat/ChatMessages/ChatMessage/ChatMessageUser/ChatMessageUserPending.svelte	Simplifies pending user message positioning
tools/ui/src/lib/components/app/chat/ChatMessages/ChatMessage/ChatMessageUser/ChatMessageUser.svelte	Shows reading stats on user message rows
tools/ui/src/lib/components/app/chat/ChatMessages/ChatMessage/ChatMessageAssistant/ChatMessageAssistant.svelte	Simplifies assistant stats view handling
tools/ui/src/lib/components/app/chat/ChatMessages/ChatMessage/ChatMessage.svelte	Wires through new user-message metadata props
tools/ui/src/lib/components/app/chat/ChatForm/ChatFormContextGauge/ContextGaugeLoadModel.svelte	New load-model prompt row
tools/ui/src/lib/components/app/chat/ChatForm/ChatFormContextGauge/ContextGaugeDial.svelte	New donut dial component
tools/ui/src/lib/components/app/chat/ChatForm/ChatFormContextGauge/ContextGaugeDetails.svelte	New hover-card details panel
tools/ui/src/lib/components/app/chat/ChatForm/ChatFormContextGauge/ContextGaugeDetailRow.svelte	New detail row component
tools/ui/src/lib/components/app/chat/ChatForm/ChatFormContextGauge/context-gauge.ts	New color/threshold helpers
tools/ui/src/lib/components/app/chat/ChatForm/ChatFormContextGauge/ChatFormContextGauge.svelte	New gauge entrypoint component
tools/ui/src/lib/components/app/chat/ChatForm/ChatFormActions/ChatFormReasoningEffortSubmenu.svelte	Adjusts submenu styling and checkmark placement
tools/ui/src/lib/components/app/chat/ChatForm/ChatFormActions/ChatFormActions.svelte	Replaces reasoning toggle with context gauge
tools/ui/src/lib/components/app/chat/ChatForm/ChatFormActions/ChatFormActionAdd/ChatFormActionAddReasoningSubmenu.svelte	Moves reasoning controls into “Add” submenu
tools/ui/src/lib/components/app/chat/ChatForm/ChatFormActions/ChatFormActionAdd/ChatFormActionAddDropdown.svelte	Adds reasoning submenu to Add dropdown
tools/ui/package.json	Bumps `@internationalized/date` version
tools/ui/package-lock.json	Locks `@internationalized/date` bump

Files not reviewed (1)

tools/ui/package-lock.json: Generated file

💡 Add Copilot custom instructions for smarter, more guided reviews. Learn how to get started.

ServeurpersoCom · 2026-07-08T06:13:57Z

Regression-free quick test PR on desktop and mobile: It remains to be seen whether this popup is readable on mobile (touch event).

ServeurpersoCom

LGTM. follow-up for mobile popup

ggerganov

Neat!

…tsVisible settings

…d live-reactive reading/output/avg

… utilities

…rmat Previously mcpStore.getToolDefinitionsForLLM() owned the MCP->OpenAI shape conversion (plus normalizeSchemaProperties). That created two sources of truth for what gets sent to the LLM, with the duplication-prone risk of the deduplicated enabled list (which feeds the token-count cache) drifting from the bytes actually shipped on chat. Now: - mcpStore: pure protocol state + routing. Drop getToolDefinitionsForLLM and the inline OpenAIToolDefinition conversion + normalizeSchemaProperties. Doc comment adjusted to declare wire-format ownership as belonging to toolsStore. Connection lifecycle, health checks, executeTool, and the connections/toolsIndex remain. - toolsStore: owns the wire shape (added earlier this series). mcpEntries() inlines the MCP tool conversion; uses normalizeJsonSchema (the JSON Schema util extracted in the prior commit) so missing 'type' fields are inferred from defaults. mcpTools getter iterates mcpEntries() so the Settings UI and the deduplicated enabled list see the same definitions. getEnabledToolsForLLM iterates mcpEntries() instead of calling mcpStore, so the JSON sent to the LLM is identical to what toolsStore.refreshEnabledToolsTokenCount tokenizes. - agentic: the chat-completion tools field's type was annotated as ReturnType<typeof mcpStore.getToolDefinitionsForLLM>, claiming the shape was owned by mcpStore. Switch to ReturnType<typeof toolsStore.getEnabledToolsForLLM>, the actual source. Assisted-by: Claude

…akdowns

…ic turn display

…-components

dagbdagb · 2026-07-08T09:48:57Z

Nice! But the visual element to trigger that panel is close to invisible...

ServeurpersoCom · 2026-07-08T10:06:31Z

Nice! But the visual element to trigger that panel is close to invisible...

This will be improved in a follow-up PR, because, as it happens, triggering this panel isn't possible on mobile for the time being.

trofen · 2026-07-11T18:37:02Z

Regarding the Reasoning control specifically: moving it from a visible button next to the input field into the “Add” menu feels less convenient and less intuitive.
Reasoning is a setting that users may toggle fairly often, so hiding it behind an extra menu makes it harder to discover and adds unnecessary friction. I think the previous placement was clearer, or at least the Reasoning toggle should remain directly accessible near the input field.

allozaur · 2026-07-14T08:37:51Z

Regarding the Reasoning control specifically: moving it from a visible button next to the input field into the “Add” menu feels less convenient and less intuitive. Reasoning is a setting that users may toggle fairly often, so hiding it behind an extra menu makes it harder to discover and adds unnecessary friction. I think the previous placement was clearer, or at least the Reasoning toggle should remain directly accessible near the input field.

hey, I'll revisit this one soon as a part of models management milestone

This makes llama-server match llama-cli's behavior of enabling reasoning by default again. Upgrading from April's llama.cpp, I thought a regression broke reasoning support. https://www.reddit.com/r/LocalLLaMA/comments/1u2yufo/cant_seem_to_enable_reasoning_in_llamacpp/ shows other people struggling with this. I tried `--reasoning on` and `--reasoning-budget -1` to no avail. Unlike Claude and Gemini which have the thinking option on the model selection button, somehow we have our reasoning limit buried in the multimedia button, where coders and text-only users would never look. I reenabled reasoning by default. Most people would want to default to using the full capabilities of the model, instead of silently restricting it. PR ggml-org#25340 changed the default from MEDIUM to OFF. This changes it back.

* feat: WIP * feat: Retire ChatScreenProcessingInfo component, context, and keepStatsVisible settings * feat: Always-on gauge with active-model /props, conversation stats and live-reactive reading/output/avg * feat: Add /tokenize endpoint, TokenizeService, FNV-1a and JSON Schema utilities * feat: Surface enabled-tools token count in context hover card * refactor(tools): make toolsStore the sole owner of the OpenAI wire format Previously mcpStore.getToolDefinitionsForLLM() owned the MCP->OpenAI shape conversion (plus normalizeSchemaProperties). That created two sources of truth for what gets sent to the LLM, with the duplication-prone risk of the deduplicated enabled list (which feeds the token-count cache) drifting from the bytes actually shipped on chat. Now: - mcpStore: pure protocol state + routing. Drop getToolDefinitionsForLLM and the inline OpenAIToolDefinition conversion + normalizeSchemaProperties. Doc comment adjusted to declare wire-format ownership as belonging to toolsStore. Connection lifecycle, health checks, executeTool, and the connections/toolsIndex remain. - toolsStore: owns the wire shape (added earlier this series). mcpEntries() inlines the MCP tool conversion; uses normalizeJsonSchema (the JSON Schema util extracted in the prior commit) so missing 'type' fields are inferred from defaults. mcpTools getter iterates mcpEntries() so the Settings UI and the deduplicated enabled list see the same definitions. getEnabledToolsForLLM iterates mcpEntries() instead of calling mcpStore, so the JSON sent to the LLM is identical to what toolsStore.refreshEnabledToolsTokenCount tokenizes. - agentic: the chat-completion tools field's type was annotated as ReturnType<typeof mcpStore.getToolDefinitionsForLLM>, claiming the shape was owned by mcpStore. Switch to ReturnType<typeof toolsStore.getEnabledToolsForLLM>, the actual source. Assisted-by: Claude * feat: UI WIP * feat: UI WIP * feat: UI WIP * feat: Adjust reasoning submenu layout and spacing * feat: Adjust context usage gauge thresholds and styling * feat: Split context usage gauge stats into current and cumulative breakdowns * chore: Format * refactor: Cleanup * refactor: Cleanup * feat: improve token gauge accuracy and display * refactor: remove MCP recommendation gating and simplify server visibility * feat: add token audit logging to ChatStore for debugging * refactor: Simplify context token reading to use server promptTokens directly * feat: Replace last-known token tracking with live server-derived stats for accurate streaming gauges * feat: UI Improvements * feat: Move prompt processing stats to the preceding user message * feat: Fix context token double-counting and refine gauge layout * refactor: remove always-show-agentic-turns setting and simplify agentic turn display * feat: track and display cache tokens in context gauge * feat: add diagnostic logging for chat completion requests * refactor: improve token audit console output with fresh/cached breakdown * fix: invalidate enabled tools token count cache on tool changes * test: add unit tests for tools store token count invalidation * refactor: Remove tools token counting infrastructure * refactor: Update ChatFormContextGauge to use simplified token tracking * refactor: Update ChatStore to remove tools token counting * chore: Formatting * feat: Improve UI text * feat: simplify context usage derivation and refine gauge labels * refactor: cleanup logs * cleaning * fix: UI * refactor: Enums * refactor: Extract context gauge logic into hook and split UI into sub-components * refactor: Cleanup comments --------- Co-authored-by: Pascal <admin@serveurperso.com>

github-actions Bot added the server/ui label Jul 6, 2026

allozaur force-pushed the feat/24346-context-usage-ui branch from e7cb827 to 6734cb6 Compare July 6, 2026 07:04

allozaur commented Jul 6, 2026

View reviewed changes

allozaur mentioned this pull request Jul 7, 2026

ui: take enabled frontend tools into account #25368

Closed

allozaur commented Jul 7, 2026

View reviewed changes

allozaur marked this pull request as ready for review July 7, 2026 20:06

allozaur requested a review from a team as a code owner July 7, 2026 20:06

allozaur requested review from ServeurpersoCom, Copilot and ggerganov July 7, 2026 20:06

Copilot started reviewing on behalf of allozaur July 7, 2026 20:07

View session

Copilot AI reviewed Jul 7, 2026

View reviewed changes

Comment thread tools/ui/src/lib/components/app/dialogs/DialogMcpServerRecommendations.svelte Resolved

Comment thread tools/ui/src/lib/components/app/content/CollapsibleContentBlock.svelte Resolved

allozaur force-pushed the feat/24346-context-usage-ui branch from fc33b19 to af49e54 Compare July 8, 2026 06:01

ServeurpersoCom approved these changes Jul 8, 2026

View reviewed changes

allozaur force-pushed the feat/24346-context-usage-ui branch from af49e54 to e7b8277 Compare July 8, 2026 07:00

ggerganov approved these changes Jul 8, 2026

View reviewed changes

allozaur added 12 commits July 8, 2026 09:15

feat: WIP

53cf38a

feat: Retire ChatScreenProcessingInfo component, context, and keepSta…

ae1e12e

…tsVisible settings

feat: Always-on gauge with active-model /props, conversation stats an…

10e2148

…d live-reactive reading/output/avg

feat: Add /tokenize endpoint, TokenizeService, FNV-1a and JSON Schema…

e979f8a

… utilities

feat: Surface enabled-tools token count in context hover card

3b88f8a

feat: UI WIP

5a2b9b8

feat: UI WIP

754dccd

feat: UI WIP

cc4a90d

feat: Adjust reasoning submenu layout and spacing

f4d0d06

feat: Adjust context usage gauge thresholds and styling

5d0197c

feat: Split context usage gauge stats into current and cumulative bre…

183dd5a

…akdowns

allozaur and others added 18 commits July 8, 2026 09:15

refactor: remove always-show-agentic-turns setting and simplify agent…

d262a1e

…ic turn display

feat: track and display cache tokens in context gauge

1b23e14

feat: add diagnostic logging for chat completion requests

9b96c22

refactor: improve token audit console output with fresh/cached breakdown

43441f6

fix: invalidate enabled tools token count cache on tool changes

7f0a850

test: add unit tests for tools store token count invalidation

3c07939

refactor: Remove tools token counting infrastructure

7f01549

refactor: Update ChatFormContextGauge to use simplified token tracking

f77b1ee

refactor: Update ChatStore to remove tools token counting

2ca0b84

chore: Formatting

c07cc64

feat: Improve UI text

5d8873d

feat: simplify context usage derivation and refine gauge labels

089095f

refactor: cleanup logs

450fb00

cleaning

d9ebe5b

fix: UI

b81b59e

refactor: Enums

2fabc94

refactor: Extract context gauge logic into hook and split UI into sub…

f34a84b

…-components

refactor: Cleanup comments

6a1d53c

allozaur force-pushed the feat/24346-context-usage-ui branch from e7b8277 to 6a1d53c Compare July 8, 2026 07:15

allozaur merged commit f1161b1 into ggml-org:master Jul 8, 2026
6 checks passed

danielzgtg mentioned this pull request Jul 21, 2026

ui: re-enable reasoning by default #25965

Closed

allozaur deleted the feat/24346-context-usage-ui branch July 29, 2026 16:46

Conversation

allozaur commented Jul 6, 2026 • edited Loading Uh oh! There was an error while loading. Please reload this page.

Uh oh!

Overview

Requirements

Uh oh!

Uh oh!

Uh oh!

Uh oh!

Uh oh!

Uh oh!

Uh oh!

Uh oh!

Uh oh!

Uh oh!

Uh oh!

ServeurpersoCom commented Jul 7, 2026

Uh oh!

Uh oh!

Uh oh!

Uh oh!

Uh oh!

Uh oh!

Copilot AI left a comment

Choose a reason for hiding this comment

Pull request overview

Reviewed changes

Uh oh!

Uh oh!

Uh oh!

ServeurpersoCom commented Jul 8, 2026

Uh oh!

ServeurpersoCom left a comment

Choose a reason for hiding this comment

Uh oh!

ggerganov left a comment

Choose a reason for hiding this comment

Uh oh!

Uh oh!

dagbdagb commented Jul 8, 2026

Uh oh!

ServeurpersoCom commented Jul 8, 2026

Uh oh!

trofen commented Jul 11, 2026

Uh oh!

allozaur commented Jul 14, 2026

Uh oh!

Reviewers

Uh oh!

Assignees

Labels

Projects

Milestone

Development

Uh oh!

6 participants

allozaur commented Jul 6, 2026 •

edited

Loading