feat: use dynamic token limits from listAvailableModels API

[chungjac]

Problem

We use hardcoded constants for LLM token/character limits.

These static values do not adapt to different model capabilities so models with larger context windows can't utilize their full capacity (Sonnet 4.5 1M), and models with smaller windows may exceed their limits.

Additionally, when users switch models mid-session, the token limits remain unchanged from the initial model selection.

Solution

Dynamically calculate character limits based on maxInputTokens from the listAvailableModels API response

Token limits are now stored per-session and recalculated when:

• Models are initially loaded
• User switches models

Updated compaction to use dynamic limits from the session

Removed deprecated static constants

Added tokenLimits to fallback models

Updated dependencies

Calculations:

• maxOverallCharacters = maxInputTokens * 3.5
• inputLimit = 0.7 * maxOverallCharacters
• compactionThreshold = 0.7 * maxOverallCharacters (leaving as separate variable from inputLimit, maybe we want to tweak these thresholds in the future)
• Default fallback: 200,000 tokens if API doesn't return tokenLimits

License

By submitting this pull request, I confirm that my contribution is made under the terms of the Apache 2.0 license.

[codecov-commenter]

Codecov Report

❌ Patch coverage is 93.75000% with 10 lines in your changes missing coverage. Please review.
✅ Project coverage is 60.16%. Comparing base (d8c5c16) to head (29b84f0).

Files with missing lines	Patch %	Lines
...nguage-server/agenticChat/agenticChatController.ts	82.45%	9 Missing and 1 partial ⚠️

Additional details and impacted files

@@            Coverage Diff             @@
##             main    #2539      +/-   ##
==========================================
+ Coverage   60.09%   60.16%   +0.07%     
==========================================
  Files         276      277       +1     
  Lines       65009    65135     +126     
  Branches     4105     4112       +7     
==========================================
+ Hits        39068    39190     +122     
- Misses      25858    25862       +4     
  Partials       83       83              

Flag	Coverage Δ
unittests	60.16% <93.75%> (+0.07%)	⬆️

Flags with carried forward coverage won't be shown. Click here to find out more.

☔ View full report in Codecov by Sentry.
📢 Have feedback on the report? Share it here.

🚀 New features to boost your workflow:

• ❄️ Test Analytics: Detect flaky tests, report on failures, and find test suite problems.
• 📦 JS Bundle Analysis: Save yourself from yourself by tracking and limiting bundle sizes in JS merges.

[rli]

seems like it would be cleaner to modify the session to encapsulate more details about the model as a single entity (i.e. model id and token limits) than to add another parameter that callers will need to remember to set

[chungjac]

Fair, will address

[chungjac]

Updated