Release list

Velune Trace v0.4.0 Latest

Latest

veluneos released this 22 Jul 09:01

v0.4.0

a7c9693

Velune Trace v0.4.0

Local Evidence and Private Baseline Preview.

Included

Local MCAP inspection
Timing profile generation
Reproducible evidence-window extraction
Versioned Core Report Bundles
Core Bundle compatibility validation
Pairwise Core Bundle comparison
Private Baseline Python APIs
Local evaluation reports
JSON source-of-truth artifacts
Bounded Markdown summaries

Release validation

316 automated tests passed
Test suite passed without a tracked sample MCAP
Test suite passed in a fresh Python virtual environment
ROS Python path was excluded from validation
Four exact dependency versions were recorded
Sample MCAP creation passed
Local Validation Report generation passed
Six required Core Report artifacts were verified

Previously measured evidence

10.7GB large-log benchmark
Approximately 9.2 million indexed events
Initial index construction measured separately from retrieval
Real-derived Core Bundle validation
Private Baseline Service E2E validation

Product boundary

Velune Trace does not automatically determine:

root cause
fault or liability
safety or severity
normality or superiority
regression or improvement

Velune Trace finds and structures evidence.
Engineers make the conclusion.

Release identity

Source commit: a7c9693531ea4c954f65419d49577b2a8d604fca
Source Tree OID: 143a26974ae10fbde800b7822d390cca05447399
Package SHA-256: d719e7d03be8a6963c3e2141b846064c56805d9b2c288dc6683bfa86ebc83ad6

Download assets

velune-trace-v0.4.0.zip — downloadable Velune Trace package
SHA256SUMS — archive SHA-256 verification record
velune-trace-v0.4.0.release-manifest.json — packaged-file integrity manifest

Verify the downloaded package before extracting it:

sha256sum -c SHA256SUMS

Continue only when the result reports:

velune-trace-v0.4.0.zip: OK

Assets 5

Velune Trace v0.3

veluneos released this 09 Jul 05:11

v0.3.6

710a747

Velune Trace v0.3 Release Notes

Release Focus

Velune Trace v0.3 is the first external validation-partner-ready release line.

It focuses on one primary workflow:

Run Velune locally on an MCAP file.
Generate ranked evidence windows.
Keep raw logs inside the user's environment.
Optionally submit one anonymous validation report.

Primary Workflow

The primary evaluation path is:

./bin/velune validation-report /path/to/your_log.mcap \
  --export-dir velune_report \
  --window-sec 1 \
  --top 5 \
  --allowed-lateness-sec 2

This creates:

velune_report/
├── summary.md
├── shareable_anonymous_report.json
├── topic_profile.json
├── evidence_windows.json
└── SCHEMA.md

Engine Proof

Velune Trace has been validated on a large robotics log benchmark:

Input size: 10.7 GB
Indexed events: 9,237,885
Rows scanned for evidence extraction: 161
Related events returned: 17
Evidence-chain retrieval after indexing: approximately 0.002 seconds
Initial index build: approximately 276 seconds
Incremental append validation: 10,000 events in approximately 0.57 seconds

Important boundary:

The 0.002 second retrieval result is measured after indexing.
Velune does not claim raw 10GB parsing in 0.002 seconds.

Validation Partner Workflow

Velune Trace can generate an anonymous report for the Validation Partner Program.

For direct validation engagements, participants may share only:

velune_report/shareable_anonymous_report.json

Raw MCAP files are not required.

Do not send:

raw MCAP files
camera images or video
LiDAR point clouds
sensor payloads
maps or location data
credentials, tokens, or keys
private operational data
customer-identifying information

Added Documentation

This release line includes:

docs/PARTNER_PROGRAM.md
docs/EXAMPLE_FEEDBACK_REPORT.md
docs/REFERENCE_COHORT_REGISTRY.md
docs/RELEASE_NOTES_v0_3_1.md
docs/RELEASE_NOTES_v0_3.md

Reference Cohort Policy

Velune must not describe a result as a global average, industry percentile, or company-to-company comparison unless a validated matched cohort exists.

Current public comparison capability:

Local evidence feedback: available
Public reference comparison: limited
Matched anonymous cohort comparison: not yet available
Global percentile dashboard: not yet available

Local-Only Trust Clarification

The CLI no longer prints a submission nudge or personal email address after generating a validation report.

Runtime output now emphasizes that no telemetry or automatic upload was performed and that all report files were written locally.

This release line adds docs/TRUST_AND_PRIVACY.md to document the local-first trust model.

Quick Start Placeholder Clarification

Quick Start examples now use /path/to/your_log.mcap instead of your_log.mcap to make it clear that users should replace the placeholder with their own MCAP file path.

Advanced CLI Commands

The following commands remain available for advanced users:

inspect
profile
windowed-verify
evidence-window

The CLI help now also presents validation-report as the first example.

For first-time evaluation, use validation-report first.

README Positioning Clarification

The README now clarifies that Velune Trace is not a Foxglove replacement.

Velune is positioned as a pre-viewer triage layer: it ranks candidate timing evidence windows first, then engineers can inspect those time ranges in Foxglove or their existing tools.

The README also clarifies the performance model:

first-pass scanning/indexing time
post-index evidence retrieval time

The benchmark numbers are described as:

10.7GB / ~9.2M-event log indexed in about 4 minutes 36 seconds
related evidence-chain query after indexing in about 0.002 seconds

The README also explains that --top can be adjusted beyond the default top 5, for example --top 10, --top 15, or --top 30.

Boundary

Velune Trace reports observable timing evidence.

Velune Trace identifies evidence windows worth reviewing first.

Velune Trace does not infer root cause.

Velune Trace does not assign fault.

Velune Trace does not assign liability.

Velune Trace does not make safety-risk determinations.

Engineers determine cause.

Public Release Name

Velune Trace v0.3

Git Tag

v0.3.6

Assets 2

Velune Trace v0.3

veluneos released this 08 Jul 16:22

v0.3.5

73ab445

Velune Trace v0.3 Release Notes

Release Focus

Velune Trace v0.3 is the first external validation-partner-ready release line.

It focuses on one primary workflow:

Run Velune locally on an MCAP file.
Generate ranked evidence windows.
Keep raw logs inside the user's environment.
Optionally submit one anonymous validation report.

Primary Workflow

The primary evaluation path is:

./bin/velune validation-report /path/to/your_log.mcap \
  --export-dir velune_report \
  --window-sec 1 \
  --top 5 \
  --allowed-lateness-sec 2

This creates:

velune_report/
├── summary.md
├── shareable_anonymous_report.json
├── topic_profile.json
├── evidence_windows.json
└── SCHEMA.md

Engine Proof

Velune Trace has been validated on a large robotics log benchmark:

Input size: 10.7 GB
Indexed events: 9,237,885
Rows scanned for evidence extraction: 161
Related events returned: 17
Evidence-chain retrieval after indexing: approximately 0.002 seconds
Initial index build: approximately 276 seconds
Incremental append validation: 10,000 events in approximately 0.57 seconds

Important boundary:

The 0.002 second retrieval result is measured after indexing.
Velune does not claim raw 10GB parsing in 0.002 seconds.

Validation Partner Workflow

Velune Trace can generate an anonymous report for the Validation Partner Program.

For direct validation engagements, participants may share only:

velune_report/shareable_anonymous_report.json

Raw MCAP files are not required.

Do not send:

raw MCAP files
camera images or video
LiDAR point clouds
sensor payloads
maps or location data
credentials, tokens, or keys
private operational data
customer-identifying information

Added Documentation

This release line includes:

docs/PARTNER_PROGRAM.md
docs/EXAMPLE_FEEDBACK_REPORT.md
docs/REFERENCE_COHORT_REGISTRY.md
docs/RELEASE_NOTES_v0_3_1.md
docs/RELEASE_NOTES_v0_3.md

Reference Cohort Policy

Velune must not describe a result as a global average, industry percentile, or company-to-company comparison unless a validated matched cohort exists.

Current public comparison capability:

Local evidence feedback: available
Public reference comparison: limited
Matched anonymous cohort comparison: not yet available
Global percentile dashboard: not yet available

Local-Only Trust Clarification

The CLI no longer prints a submission nudge or personal email address after generating a validation report.

Runtime output now emphasizes that no telemetry or automatic upload was performed and that all report files were written locally.

This release line adds docs/TRUST_AND_PRIVACY.md to document the local-first trust model.

Quick Start Placeholder Clarification

Quick Start examples now use /path/to/your_log.mcap instead of your_log.mcap to make it clear that users should replace the placeholder with their own MCAP file path.

Advanced CLI Commands

The following commands remain available for advanced users:

inspect
profile
windowed-verify
evidence-window

The CLI help now also presents validation-report as the first example.

For first-time evaluation, use validation-report first.

Boundary

Velune Trace reports observable timing evidence.

Velune Trace identifies evidence windows worth reviewing first.

Velune Trace does not infer root cause.

Velune Trace does not assign fault.

Velune Trace does not assign liability.

Velune Trace does not make safety-risk determinations.

Engineers determine cause.

Public Release Name

Velune Trace v0.3

Git Tag

v0.3.5

Assets 2

Velune Trace v0.3

veluneos released this 08 Jul 08:50

v0.3.4

2502aff

Velune Trace v0.3 Release Notes

Release Focus

Velune Trace v0.3 is the first external validation-partner-ready release line.

It focuses on one primary workflow:

Run Velune locally on an MCAP file.
Generate ranked evidence windows.
Keep raw logs inside the user's environment.
Optionally submit one anonymous validation report.

Primary Workflow

The primary evaluation path is:

./bin/velune validation-report /path/to/your_log.mcap \
  --export-dir velune_report \
  --window-sec 1 \
  --top 5 \
  --allowed-lateness-sec 2

This creates:

velune_report/
├── summary.md
├── shareable_anonymous_report.json
├── topic_profile.json
├── evidence_windows.json
└── SCHEMA.md

Engine Proof

Velune Trace has been validated on a large robotics log benchmark:

Input size: 10.7 GB
Indexed events: 9,237,885
Rows scanned for evidence extraction: 161
Related events returned: 17
Evidence-chain retrieval after indexing: approximately 0.002 seconds
Initial index build: approximately 276 seconds
Incremental append validation: 10,000 events in approximately 0.57 seconds

Important boundary:

The 0.002 second retrieval result is measured after indexing.
Velune does not claim raw 10GB parsing in 0.002 seconds.

Validation Partner Workflow

Velune Trace can generate an anonymous report for the Validation Partner Program.

Participants may submit only:

velune_report/shareable_anonymous_report.json

Raw MCAP files are not required.

Do not send:

raw MCAP files
camera images or video
LiDAR point clouds
sensor payloads
maps or location data
credentials, tokens, or keys
private operational data
customer-identifying information

Added Documentation

This release line includes:

docs/PARTNER_PROGRAM.md
docs/EXAMPLE_FEEDBACK_REPORT.md
docs/REFERENCE_COHORT_REGISTRY.md
docs/RELEASE_NOTES_v0_3_1.md
docs/RELEASE_NOTES_v0_3.md

Reference Cohort Policy

Velune must not describe a result as a global average, industry percentile, or company-to-company comparison unless a validated matched cohort exists.

Current public comparison capability:

Local evidence feedback: available
Public reference comparison: limited
Matched anonymous cohort comparison: not yet available
Global percentile dashboard: not yet available

Quick Start Placeholder Clarification

Quick Start examples now use /path/to/your_log.mcap instead of your_log.mcap to make it clear that users should replace the placeholder with their own MCAP file path.

Advanced CLI Commands

The following commands remain available for advanced users:

inspect
profile
windowed-verify
evidence-window

The CLI help now also presents validation-report as the first example.

For first-time evaluation, use validation-report first.

Boundary

Velune Trace reports observable timing evidence.

Velune Trace identifies evidence windows worth reviewing first.

Velune Trace does not infer root cause.

Velune Trace does not assign fault.

Velune Trace does not assign liability.

Velune Trace does not make safety-risk determinations.

Engineers determine cause.

Public Release Name

Velune Trace v0.3

Git Tag

v0.3.4

Assets 2

Velune Trace v0.3

veluneos released this 08 Jul 08:25

v0.3.3

3ec38fc

Velune Trace v0.3 Release Notes

Release Focus

Velune Trace v0.3 is the first external validation-partner-ready release line.

It focuses on one primary workflow:

Run Velune locally on an MCAP file.
Generate ranked evidence windows.
Keep raw logs inside the user's environment.
Optionally submit one anonymous validation report.

Primary Workflow

The primary evaluation path is:

./bin/velune validation-report your_log.mcap \
  --export-dir velune_report \
  --window-sec 1 \
  --top 5 \
  --allowed-lateness-sec 2

This creates:

velune_report/
├── summary.md
├── shareable_anonymous_report.json
├── topic_profile.json
├── evidence_windows.json
└── SCHEMA.md

Engine Proof

Velune Trace has been validated on a large robotics log benchmark:

Input size: 10.7 GB
Indexed events: 9,237,885
Rows scanned for evidence extraction: 161
Related events returned: 17
Evidence-chain retrieval after indexing: approximately 0.002 seconds
Initial index build: approximately 276 seconds
Incremental append validation: 10,000 events in approximately 0.57 seconds

Important boundary:

The 0.002 second retrieval result is measured after indexing.
Velune does not claim raw 10GB parsing in 0.002 seconds.

Validation Partner Workflow

Velune Trace can generate an anonymous report for the Validation Partner Program.

Participants may submit only:

velune_report/shareable_anonymous_report.json

Raw MCAP files are not required.

Do not send:

raw MCAP files
camera images or video
LiDAR point clouds
sensor payloads
maps or location data
credentials, tokens, or keys
private operational data
customer-identifying information

Added Documentation

This release line includes:

docs/PARTNER_PROGRAM.md
docs/EXAMPLE_FEEDBACK_REPORT.md
docs/REFERENCE_COHORT_REGISTRY.md
docs/RELEASE_NOTES_v0_3_1.md
docs/RELEASE_NOTES_v0_3.md

Reference Cohort Policy

Velune must not describe a result as a global average, industry percentile, or company-to-company comparison unless a validated matched cohort exists.

Current public comparison capability:

Local evidence feedback: available
Public reference comparison: limited
Matched anonymous cohort comparison: not yet available
Global percentile dashboard: not yet available

Advanced CLI Commands

The following commands remain available for advanced users:

inspect
profile
windowed-verify
evidence-window

The CLI help now also presents validation-report as the first example.

For first-time evaluation, use validation-report first.

Boundary

Velune Trace reports observable timing evidence.

Velune Trace identifies evidence windows worth reviewing first.

Velune Trace does not infer root cause.

Velune Trace does not assign fault.

Velune Trace does not assign liability.

Velune Trace does not make safety-risk determinations.

Engineers determine cause.

Public Release Name

Velune Trace v0.3

Git Tag

v0.3.3

Assets 2

Velune Trace v0.3

veluneos released this 08 Jul 08:16

v0.3.2

fd3c1f1

Velune Trace v0.3 Release Notes

Release Focus

Velune Trace v0.3 is the first external validation-partner-ready release line.

It focuses on one primary workflow:

Run Velune locally on an MCAP file.
Generate ranked evidence windows.
Keep raw logs inside the user's environment.
Optionally submit one anonymous validation report.

Primary Workflow

The primary evaluation path is:

./bin/velune validation-report your_log.mcap \
  --export-dir velune_report \
  --window-sec 1 \
  --top 5 \
  --allowed-lateness-sec 2

This creates:

velune_report/
├── summary.md
├── shareable_anonymous_report.json
├── topic_profile.json
├── evidence_windows.json
└── SCHEMA.md

Engine Proof

Velune Trace has been validated on a large robotics log benchmark:

Input size: 10.7 GB
Indexed events: 9,237,885
Rows scanned for evidence extraction: 161
Related events returned: 17
Evidence-chain retrieval after indexing: approximately 0.002 seconds
Initial index build: approximately 276 seconds
Incremental append validation: 10,000 events in approximately 0.57 seconds

Important boundary:

The 0.002 second retrieval result is measured after indexing.
Velune does not claim raw 10GB parsing in 0.002 seconds.

Validation Partner Workflow

Velune Trace can generate an anonymous report for the Validation Partner Program.

Participants may submit only:

velune_report/shareable_anonymous_report.json

Raw MCAP files are not required.

Do not send:

raw MCAP files
camera images or video
LiDAR point clouds
sensor payloads
maps or location data
credentials, tokens, or keys
private operational data
customer-identifying information

Added Documentation

This release line includes:

docs/PARTNER_PROGRAM.md
docs/EXAMPLE_FEEDBACK_REPORT.md
docs/REFERENCE_COHORT_REGISTRY.md
docs/RELEASE_NOTES_v0_3_1.md
docs/RELEASE_NOTES_v0_3.md

Reference Cohort Policy

Velune must not describe a result as a global average, industry percentile, or company-to-company comparison unless a validated matched cohort exists.

Current public comparison capability:

Local evidence feedback: available
Public reference comparison: limited
Matched anonymous cohort comparison: not yet available
Global percentile dashboard: not yet available

Advanced CLI Commands

The following commands remain available for advanced users:

inspect
profile
windowed-verify
evidence-window

For first-time evaluation, use validation-report first.

Boundary

Velune Trace reports observable timing evidence.

Velune Trace identifies evidence windows worth reviewing first.

Velune Trace does not infer root cause.

Velune Trace does not assign fault.

Velune Trace does not assign liability.

Velune Trace does not make safety-risk determinations.

Engineers determine cause.

Public Release Name

Velune Trace v0.3

Git Tag

v0.3.2

Assets 2

Velune Trace v0.3.1

veluneos released this 08 Jul 07:57

v0.3.1

7d3c80e

Velune Trace v0.3.1 Release Notes

Release Focus

v0.3.1 focuses on making Velune Trace easier for external robotics engineers to run, evaluate, and use as a local evidence-window extraction engine.

This release highlights the large-log evidence extraction proof and adds the Validation Partner workflow.

Main Additions

1. Validation Report Workflow

Added the validation-report command as the primary external evaluation path.

Users can run Velune locally on an MCAP file and generate:

summary.md
evidence_windows.json
topic_profile.json
shareable_anonymous_report.json
SCHEMA.md

Raw MCAP files are not required for partner submission.

2. Anonymous Report Submission

Added a shareable anonymous report path for the Velune Validation Partner Program.

Participants may submit only:

velune_report/shareable_anonymous_report.json

The report is designed to avoid raw sensor payloads.

3. Partner Onboarding Documentation

Added:

docs/PARTNER_PROGRAM.md
docs/EXAMPLE_FEEDBACK_REPORT.md
docs/REFERENCE_COHORT_REGISTRY.md

These documents explain what participants run, what they may share, what they must not share, and what kind of feedback they may receive.

4. Reference Cohort Policy

Added the Reference Cohort Registry policy to prevent overclaiming.

The policy separates:

synthetic samples
public reference datasets
internal benchmark corpora
controlled fixtures
future partner anonymous reports

Velune must not claim global averages, industry percentiles, or company-to-company comparisons unless a validated matched cohort exists.

5. Large-Log Engine Proof in README

Updated the README to emphasize Velune Trace as an evidence-window extraction engine for large robotics logs.

Highlighted validation evidence:

10.7 GB benchmark corpus
9,237,885 indexed events
161 rows scanned for evidence extraction
17 related events returned
approximately 0.002 second evidence-chain retrieval after indexing
approximately 276 second initial index build
10,000 event incremental append validation in approximately 0.57 seconds

Boundary

Velune Trace reports observable timing evidence.

Velune Trace identifies evidence windows worth reviewing first.

Velune Trace does not infer root cause.

Velune Trace does not assign fault.

Velune Trace does not assign liability.

Velune Trace does not make safety-risk determinations.

Engineers determine cause.

Recommended Tag

v0.3.1

Assets 2

v0.2.0

veluneos released this 13 Jun 12:01

v0.2.0

72a718a

Initial public release of Velune Trace.

Highlights

10GB ROS2-style trace corpus benchmark
9.2M indexed events
Event-chain retrieval after indexing
Append-only incremental indexing
Failure-mode validation
Streaming trace viewer

Validation Snapshot

Metric	Result
Corpus Size	10 GB
Events Indexed	9,237,885
Initial Index Build	276 sec
SQLite Index Size	~4.8 GB
Event Chain Retrieval	0.002 sec
Incremental Append	10,000 events in 0.57 sec

Current Scope

Validated on:

ROS2-style structured JSONL trace logs
SQLite indexing
Deterministic event-chain reconstruction
Append-only incremental indexing

Not yet validated on:

100GB+ corpora
1TB-scale fleet data
Distributed indexing
Multi-user deployments
Unstructured legacy logs
Real-time monitoring workloads

Non-Claims

Velune Trace does not perform:

Root cause analysis
Fault attribution
Liability assignment
Risk scoring
Judgment engines

Find the events. Engineers find the cause.

Assets 2

Releases: veluneos/velune-trace

Release list