🔐 ORDS OAuth Token Retrieval Utility, Updated Examples & Improved Diagnostics and Validation

✨ Added

• Added get_vecdb_oauth_token for OAuth client-credentials requests, with HTTPS validation, configurable TLS, timeouts and proxies, and sanitized OAuthTokenError diagnostics.
• Added network failure details to VecDBException, including whether the failure involved a timeout, DNS lookup, connection, or TLS/certificate issue, and whether retrying may help.

🔄 Changed

• Network failures now include distinct error codes and guidance on what to check next.
• query.top_k now requires a positive integer; booleans and fractional values are rejected.
• Removed distribute_method validation when distribute_params is null.

🐛 Fixed

• Made exception formatting handle unsupported, circular, and deeply nested diagnostic values while keeping sensitive data redacted.

📦 Installation

python -m pip install --upgrade oracle-vecdb

Full Changelog: v1.0.5...v1.0.6

⏱️ Configurable Timeouts, Improved Retries & Clearer Errors

Release 1.0.5 improves HTTP request control, response handling, retry behavior, and diagnostics for Oracle VecDB Python SDK applications.

Added

• Added configurable HTTP request timeouts through Configuration.timeout, supporting total timeouts and separate connect/read timeouts.
• Added direct iteration support for query, paged, and vector collection responses.
• Added configurable retry-delay limits for transient HTTP 429 responses, including support for server-provided Retry-After values.

Fixed

• Improved upsert error messages for missing or oversized vector IDs, including guidance on enabling automatic ID generation.
• Removed REST endpoint values from being echoed in URL validation errors.

Install

python -m pip install --upgrade oracle-vecdb

Full Changelog: v1.0.4...v1.0.5

This release strengthens exception diagnostics and streamlines the project’s
documentation sources.

✨ Added

• Added exception diagnostic redaction for sensitive values in nested
  arguments.

🔄 Changed

• Removed the obsolete Sphinx documentation tree; the repository README and
  published reference sources remain the documentation entry points.

📦 Installation

python -m pip install --upgrade oracle-vecdb

See the full changelog or compare v1.0.3 to v1.0.4.

This patch release improves client-side validation, default handling, authentication configuration, upsert compatibility, and ORDS error handling.

✨ Added

• Added consistent defaults for common operations, including table creation, vector listing, and query responses.
• Added client-side validation for parameter values and incompatible parameter combinations before requests are sent to ORDS.
• Added explicit validation for incomplete or conflicting authentication settings.

🔄 Changed

• Upsert vector field names are now matched case-insensitively, with clear errors for unknown or duplicate fields.
• Improved resource-name validation and error messages while preserving database-specific identifier rules.
• Improved ORDS exception handling to redact sensitive request and response data and avoid exposing duplicate raw transport details.

🐛 Fixed

• Fixed client-side validation for HNSW (INMEMORY GRAPH) index requests.

📦 Installation

python -m pip install --upgrade oracle-vecdb

See the full changelog or compare v1.0.2 to v1.0.3.

🌍 Geospatial Search, 💻 Code Search & 🤖 RAG Examples

This documentation release makes it easier to go from the Oracle VecDB Python SDK to runnable examples and notebooks in the companion Oracle AI Developer Hub.

Explore how Oracle VecDB can be used for geospatial + semantic search, semantic code search, RAG document chat, and more.

🌍 Geospatial + Semantic Search

Find what you mean, where you need it.

Ask the Parks combines semantic vector search with city/radius and GeoJSON filtering over National Park Service data.

➡️ Explore Ask the Parks

💻 Semantic Code Search

Search functions, classes, and source code using natural-language intent rather than exact keywords.

➡️ Explore Semantic Code Search

🤖 RAG Document Chat

Build an end-to-end RAG workflow using Oracle VecDB as the retrieval layer, with support for local Ollama models and OpenAI-compatible APIs.

➡️ Explore the Doc Chatbot

📓 Runnable Notebooks

Prefer to learn interactively? The companion notebook collection includes hands-on workflows for embeddings, vector ingestion, similarity search, metadata filtering, HNSW tuning, RAG, and more.

➡️ Explore the VecDB notebooks

Changed

• Refreshed the README to make runnable VecDB applications and notebooks easier to discover.
• Added direct paths from the SDK repository to geospatial search, semantic code search, and RAG examples.
• Improved the getting-started journey from SDK installation to hands-on examples.

Note: v1.0.2 contains documentation changes only. There are no SDK API or runtime changes in this release.

Install

python -m pip install --upgrade oracle-vecdb

Changed

• Updated the README quickstart to align with the published Python SDK documentation.
• Added clearer installation and configuration guidance.
• Documented integrated embedding workflows.
• Added bring-your-own-vectors (BYOV) examples.
• Added filtered query examples.
• Expanded index creation and query-tuning guidance.
• Updated the SDK sample client with table lifecycle, metadata, listing, filtering, deletion, and cleanup examples.
• Updated the package homepage URL to the current Oracle documentation.

Full Changelog: v1.0.0...v1.0.1

oracle-vecdb 1.0.0 is now released. This is the initial release and adds support for vector table management, indexing, search, and inference operations in Oracle AI Database 23.26.3+ using ORDS 26.2.2+ API's.