V0.0.39.2

Changes

Increase emissions control percent to manage supply / stabilize EMA.

v0.0.39.1

Release Notes - v0.0.39.1

Overview

This release focuses on improving snippet fetching reliability and test coverage. Key changes include HTTP header optimization, simplified response handling, and enhanced test infrastructure.

Changes

🔧 Snippet Fetcher Improvements

HTTP Header Optimization

• Updated Accept-Encoding header: Removed Brotli (br) compression support, keeping only gzip and deflate
  • Reason: Some servers may not handle Brotli compression correctly, causing binary data issues
  • Impact: More reliable decompression and text extraction from web responses
  • Files Modified: validator/snippet_fetcher.py

Response Handling Simplification

• Removed content type validation checks: Simplified response processing by removing binary data detection and content-type validation
  • Reason: httpx automatically handles decompression and encoding, making manual checks redundant
  • Impact: Cleaner code, faster processing, reduced false positives
  • Files Modified: validator/snippet_fetcher.py
  • Lines Removed: ~45 lines of validation code

🧪 Test Infrastructure Updates

Test Suite Improvements

• Removed manual test file: Deleted tests/manual/test_snippet_fetcher.py (136 lines removed)
• Enhanced unit tests: Updated tests/unit_tests/test_snippet_fetcher.py with comprehensive improvements
  • Expanded test URL coverage with 60 URLs from 20 different domains
  • Test URLs extracted from presentational_df_copy.csv (3 samples per domain)
  • Added logging support for better test debugging
  • Improved test organization and structure

New Test Domains

Added test coverage for the following domains:

• animalsaroundtheglobe.com
• biologyinsights.com
• businessinsider.com
• cnn.com
• encyclopedia.com
• facts.net
• fandom.com (multiple subdomains)
• fastercapital.com
• fiveable.me
• indiatimes.com
• livescience.com
• medium.com
• nasa.gov
• space.com
• studysmarter.co.uk
• testbook.com
• thecollector.com
• usatoday.com
• vaia.com
• vocal.media

Breaking Changes

None - This is a patch release with backward-compatible improvements.

Migration Notes

No migration required. The changes are internal improvements that don't affect the API or external interfaces.

Testing

• Enhanced test coverage with 60 URLs across 20 domains
• Improved logging for test debugging

Fixed encoding when fetching html snippets.

v0.0.39

🚀 Release Notes – v0.0.39

This release includes reliability upgrades and a major expansion of the evaluation rubric used across the system.

🔧 Improvements

• Implemented a fallback HTML fetch mechanism to ensure more consistent snippet retrieval when the primary method fails.

🧠 Enhanced Evaluation Rubric

The rubric now captures deeper signal quality by scoring six new dimensions:

• Sentiment (sentiment, -1.0 → 1.0)
  Measures positive vs negative tone toward the statement.

• Conviction (conviction, 0.0 → 1.0)
  Estimates how confident and certain the response is.

• Source Credibility (source_credibility, 0.0 → 1.0)
  Evaluates the trustworthiness and authority of cited or implied sources.

• Narrative Momentum (narrative_momentum, 0.0 → 1.0)
  Scores how well the explanation builds a coherent, engaging narrative.

• Risk–Reward Sentiment (risk_reward_sentiment, -1.0 → 1.0)
  Captures whether the reasoning emphasizes risks (negative) or rewards (positive).

• Catalyst Detection (catalyst_detection, 0.0 → 1.0)
  Measures how effectively the model identifies key triggers, events, or drivers related to the statement.

📦 Summary

A stability-focused release with expanded rubric dimensions to support richer miner scoring and better signal evaluation.

v0.0.38

Release v0.0.38

Summary

• Weight calculation: added ranking- and exponential-based weight distribution methods, refactored emission-control (burn) logic, and centralized weight conversion.
• SnippetFetcher: made requests look more browser-like (rotating User-Agents, realistic headers, cookie jar, referer support) and improved error handling/logging.
• Domain validation: fixed timezone / naive-datetime comparisons to avoid incorrect domain-age checks.
• Misc: refresh interval reduced, IMMUNITY_PERIOD tweaked, .gitignore updates, and new unit + manual tests.

Key changes

validator/validator_daemon.py

• New constants: DEFAULT_TOTAL_WEIGHT, EXPONENTIAL_DECAY_SCALE, USE_RANKING_EMISSION_CONTROL, RANKING_EMISSION_TOP_PERC
• Refactor and validation helpers: find_burn_miner(...) and burn_weights(...)
• New distribution functions:
  • distribute_weights_by_ranking(...) — geometric progression by rank (top gets configurable percent, next gets half, etc.)
  • distribute_weights_by_exponential_decay(...) — smooth exponential-decay distribution
  • convert_scores_to_weights(...) and move_miner_weights(...) — centralize conversion and optional emission burning
• Main loop now calls move_miner_weights(...)

validator/snippet_fetcher.py

• USER_AGENTS rotation and _get_browser_headers(...) to generate realistic Accept / Sec-Fetch / Referer headers
• AsyncClient now uses httpx.Cookies() (cookie jar)
• send_get_request(...) merges realistic headers, accepts referer, logs 403/429, and handles httpx.TimeoutException / HTTPStatusError / RequestError / generic Exception (returns None on failure)
• render_page(...) forwards referer
• Manual test harness updated to support multiple test URLs and CLI flags (--all, --list)

validator/domain_validator.py

• Use timezone-aware now (datetime.now(timezone.utc))
• If creation_date is naive, assume UTC (creation_date = creation_date.replace(tzinfo=timezone.utc)) before comparing age

validator/api_server.py

• REFRESH_INTERVAL_SECONDS changed from 20 minutes → 10 minutes

shared/environment_variables.py

• IMMUNITY_PERIOD changed from 350 → 100 (review for intended units/semantics)

v0.0.37

Release v0.0.37 — Summary

• Added version endpoint

  • Introduces a new endpoint that exposes the service/API version for consumers and health checks.

• Upgraded logging to include validator UID

  • Logging now includes the validator unique identifier (UID) with relevant events to improve traceability and debugging.

• Added additional timing to validator → miner responses

  • Captures extra timing metrics on validator-to-miner response flows to help diagnose latency and performance issues.

v0.0.36

Release v0.0.36

🔧 Scoring Sensitivity Adjustment

We refined the score weighting mechanism by reducing the exponential decay scale from 8.0 to 4.0.
This change increases the sensitivity of the weighting function, making higher scores more dominant while reducing the relative influence of lower scores.

Impact

• The exponential decay now reacts more sharply to score differences.
• Top-performing scores receive proportionally higher weights.
• Overall weighting distribution becomes tighter and more selective.

Reasoning

This adjustment was made to enhance differentiation among closely ranked scores, ensuring that small performance improvements are more accurately reflected in the final weighting process.

v0.0.35

Release v0.0.35

Purpose

Support the price of our alpha token in a way that is competitive with other subnets.

Background

Many subnets are using emission control mechanisms to reduce effective inflation / sell pressure and support the price of their alpha token. Some are even burning 100%! Check out: https://huggingface.co/spaces/pawkanarek/burntensor

Implementation

The Vericore team will manage a separate wallet associated with this emission control mechanism. A portion of the rewards will go to that wallet once it is registered. We plan to use the wallet to lock up tokens, burn tokens, and disperse rewards to top miners after the rewards have been held for a period. Exact details TBD - we will make announcements in the subnet discord.

Changes

These changes update the validator daemon by adding code that finds the Vericore team's emissions control wallet in the metagraph and sends a percentage of emissions to that wallet.

Details

• updates validator/validator_daemon.py with burn logic
• emission percent hardcoded to 50%

v0.0.34

Release v0.0.34

Changes

These changes collectively update how miner weights are handled, introduce an immunity period for new miners, and refactor some configuration aspects to use constants, along with other minor improvements.

v0.0.33

Release v0.0.33

Changes

• Switched off Html Parser API

v0.0.32

Release v0.0.32

Changes

• Reverted change to api_server