Skip to main content

Staqtapp-TDS v3.11.0

Staqtapp-TDS 3.11.0 is the final public version of TDS.

Platform: Linux.

This public distribution does not include Eaglegate activation or full Eaglegate integration. The Olum multi-agentic registry system and workflow are not included.

Eaglegate target and shadow qualification utilities remain non-authoritative and cannot activate production serving.

Staqtapp-TDS is a directory-first storage, persistence, evidence, and operations engine for Python systems. It stores typed values in an indexed hierarchy, persists exact bytes to rooted .tds files, and returns stable non-halting results for ordinary operations.

Programmer Core API Guide (PDF; complete source ZIP)

Package status Value
Version 3.11.0
Python 3.10+
Platform Linux
Base dependency NumPy >=1.23
License MIT

Available features

Area Included
Directory storage Hierarchical directories; exact indexed lookup; binary, text, JSON, NumPy, and restricted-pickle values; schemas; bounded batch reads and writes; locks; and stalk chains.
Result-first API TDSResult responses with stable codes, values, messages, and metadata for storage, variable, persistence, and capability operations.
Rooted .tds persistence Atomic snapshots, exact encoded-byte retention, SHA-256 roots, required integrity sidecars for new v2 files, lazy mounts, immutable reader epochs, pin/reload safety, and integrity audits.
Versioned storage Self-contained .tds v3 nodes, dirty-node Merkle commits, frozen indexes, NONE and ZLIB blocks, leased range reads, read-only NumPy access, and cross-process publication locks.
Generations and CAS SHA-256 sharded content-addressed storage, immutable publications, head compare-and-swap, reader pins, recovery, rollback, retirement, and composite CSV objects.
CSV evidence Original-byte preservation, dialect fingerprints, row offsets and anchors, scans, validation, artifact transactions, storage binding, Interpole evidence, and explicit Semantic IR review records.
Trace and paths Deterministic trace ranking, packed waypoint/CSR graphs, provenance records, bounded path planning, and replayable receipts.
Driver platform TDDL validation, deterministic bytecode, bounded Driver VM execution, Runtime Manager evidence, Foundry proposal/review flows, regression harnesses, audit exports, and the optional PySide6 Stargate cockpit.
Operations Cached telemetry, bounded diagnostics, pressure/recovery reporting, local Stargate Browser console, CLI verification, and optional workspace snapshots.
Acceleration Optional Numba paths and opt-in C index/CSV kernels with deterministic Python fallbacks.

Measured performance

PyPI 3.8.3 compared with 3.11.0

Workload 3.8.3 median 3.11.0 median Improvement
Exact hit, telemetry OFF 9.125 µs/read 1.253 µs/read 7.28×
Exact hit, telemetry LIGHT 13.824 µs/read 3.405 µs/read 4.06×
Exact miss, telemetry OFF 2.894 µs/read 1.347 µs/read 2.15×
Exact miss, telemetry LIGHT 7.747 µs/read 3.662 µs/read 2.12×
Probability-ranked scan, 2,047 entries 9.632 µs/read 1.937 µs/read 4.97×
Probability-ranked scan, 2,049 entries 1.934 ms/read 1.912 µs/read 1,011.31×
Probability-ranked scan, 4,096 entries 1.915 ms/read 2.271 µs/read 843.34×
Portable slot packing, 8,192 records 15.038 µs/record 0.505 µs/record 29.80×

The 843–1,011× results are the removal of the old post-2,048-entry registry scaling and correctness cliff, not general TDS throughput multipliers. Version 3.8.3 omitted live keys in those fixtures; 3.11.0 retained exact closure.

Additional 3.11.0 medians

Workload Result
Rooted v2 flush, 64 MiB uncompressed 210.645 ms; 303.8 MiB/s; 1,070,551 B traced Python peak
Telemetry record_read 1,441,534 calls/s
Canonical JSON dump / fast load 2.744 / 1.956 µs per operation
CSV scan plus row anchors, 20,001 rows 85.620 ms
CSV import, same fixture 67.265 ms
CSV row-offset packing 3.492 ms; 320,463 B traced Python peak
Eight cached root reads, 4,096 payloads 0.004817 ms
Direct / managed Driver VM, 5,000 records 93.983 / 204.914 ms
Trace graph / packed-byte admission, 1,600 rows 4.464 / 6.186 ms

Benchmark scope:

  • Linux 6.18.35 x86-64, one AMD EPYC 9V74 core, CPython 3.12.13, NumPy 2.3.5.
  • Numba disabled, CPU affinity 4, PYTHONHASHSEED=0, two warmups, nine timed repetitions.
  • Paired results were accepted only after values, telemetry, packed bytes, round trips, and SHA-256 roots matched.
  • Results are workload-specific local measurements. tracemalloc is not process RSS.

Benchmark harnesses are included in the complete source ZIP.

Install

# Base package
python -m pip install "staqtapp-tds==3.11.0"

# Optional Numba acceleration
python -m pip install "staqtapp-tds[fast]==3.11.0"

# Optional PySide6 cockpit
python -m pip install "staqtapp-tds[gui]==3.11.0"

The PyPI wheel uses deterministic Python backends. Optional native builds are Linux-only and require an explicit source build:

STAQTAPP_TDS_BUILD_NATIVE=1 \
  python -m pip install --no-binary staqtapp-tds "staqtapp-tds==3.11.0"

Quick start

from pathlib import Path

from staqtapp_tds import TDSFileSystem, TDSPersistence, TDSReader

fs = TDSFileSystem("agent_state")
runtime = fs.makedirs("/models/runtime")

assert runtime.write_text("system_prompt", "Be precise.").ok
assert runtime.write_json("settings", {"temperature": 0.2}).ok
assert runtime.write_result("step_count", 7).ok

result = runtime.read_result("settings")
assert result.ok and result.value["temperature"] == 0.2

store = TDSPersistence(Path("./tds_store"))
store.flush(fs, parallel_nodes=False)

path = Path("./tds_store/agent_state__models__runtime.tds")
assert store.load_node(path).read_value("step_count") == 7

with TDSReader(path) as reader:
    assert reader.audit_integrity()["ok"]

Commands

Command Purpose
staqtapp-tds Start the local telemetry Browser on 127.0.0.1:8765.
staqtapp-tds-admin status Show installed runtime status.
staqtapp-tds-admin verify --sample Run packaged health and sample-data checks.
staqtapp-tds-generation-audit Audit immutable-generation publication, pins, rollback, and recovery.
staqtapp-tds-driver-foundry Open the optional PySide6 cockpit.

Format and authority boundaries

Boundary Status
Package and formats Package 3.11.0, .tds v2, additive .tds v3, Generation v1, CSV generation v1, and tds-csv-composite-v2 are separate identities.
Existing data Version 3.11.0 retains existing .tds v2 bytes and layout.
Encryption At-rest encryption is not implemented; DirFlags.ENCRYPTED fails closed. XorCryptoProvider is not secure encryption.
Input trust Treat .tds and pickle-compatible inputs as trusted until application-specific resource and trust controls are applied.
Semantic authority TDS records caller declarations and authorized review transitions; it does not infer or commit semantic truth.
Driver authority Driver Foundry and Studio do not sign or activate drivers.
Eaglegate Activation and full integration are not included.
Olum The multi-agentic registry system and workflow are not included.

License

MIT. See LICENSE.

Release files for staqtapp-tds 3.11.0

For a detailed explanation of source distributions (sdists) and built distributions (wheels), please see the package formats documentation.

Source distribution (sdist)

Source distribution for staqtapp-tds 3.11.0
File Size Uploaded
staqtapp_tds-3.11.0.tar.gz 739.3 kB Details

Built distribution (wheel)

Table of built distributions (wheels) for staqtapp-tds 3.11.0
File Interpreter ABI Platform
staqtapp_tds-3.11.0-py3-none-any.whl Python 3 none any Details

Total release size: 1.6 MB

Release files / staqtapp_tds-3.11.0.tar.gz

Download URL staqtapp_tds-3.11.0.tar.gz
Size 739.3 kB
Tags Source
SHA-256 checksum
How to use checksums
d5c388891bdfb936017c4393bdf73c0385cd8eba0d635fb02247e22388c4fe27
BLAKE2b-256 checksum
How to use checksums
83e7abb9d0b2698716b8f680bb25accc4a0f15cd4a7323f10e608c092b110f38
Upload date
Uploaded using Trusted Publishing?
What is trusted publishing?
No
Uploaded via twine/0.0.0 importlib_metadata/4.0.1 pkginfo/1.12.1.2 requests/2.25.1 requests-toolbelt/0.9.1 tqdm/4.61.1 CPython/3.9.16

Release files / staqtapp_tds-3.11.0-py3-none-any.whl

Download URL staqtapp_tds-3.11.0-py3-none-any.whl
Size 847.0 kB
Tags Python 3
SHA-256 checksum
How to use checksums
907784b959715f573841dd423de2e43b5edcabbb0acec65d894341bdb8aa2eeb
BLAKE2b-256 checksum
How to use checksums
a4661c3b96554c2601f107d6723a8fa1a3cb840f989e2c055e292886aa0c2232
Upload date
Uploaded using Trusted Publishing?
What is trusted publishing?
No
Uploaded via twine/0.0.0 importlib_metadata/4.0.1 pkginfo/1.12.1.2 requests/2.25.1 requests-toolbelt/0.9.1 tqdm/4.61.1 CPython/3.9.16

Release history Release notifications | RSS feed

This release

3.11.0 This release

2 release files

3.8.3

2 release files

3.8.2

2 release files

3.8.1

2 release files

3.8.0

2 release files

3.5.3

2 release files

Anthropic, PBC Visionary sponsor Bloomberg Visionary sponsor Hudson River Trading Visionary sponsor Meta Visionary sponsor NVIDIA Visionary sponsor Microsoft Sustainability sponsor Depot Continuous Integration AWS Cloud computing and Security Sponsor Datadog Monitoring Fastly CDN Google Download Analytics Sentry Error logging StatusPage Status page