Product Docs Weekly — issue 001, August 26, 2026

New Enterprise model controls let admins choose which Agent model providers and models are available in each Enterprise Workspace.

New SSO integration via the Clerk dashboard lets app users sign in with company identity providers (Okta, Microsoft Entra ID) plus multi-factor authentication and session controls.

New SAML SSO setup for Replit Enterprise Organizations is now configurable directly from workspace settings, including IdP configuration and email domain claims.

New Project Security Center lets developers choose a security scan, review findings, and fix vulnerabilities before publishing a Replit App.

New agent security scans enable source-code security review, finding assessment, Agent-assisted remediation, and fix verification in a single workflow.

New black-box pen tests (Level 3 security scan) combine source-code review with live browser and network interaction testing against the running app.

New Steer and Queue controls let users redirect or schedule follow-up messages to Agent mid-task, with Conversation and Project-level preferences.

Adds new shell completions for mpv, ruff, deno, yay, paru, kubectl --user value completions, ip (address, link, route, neighbour, netns, rule, tunnel, xfrm, monitor, and remaining subcommands), and git read-tree.

C# and TypeScript enum members now each emit a graph node with a case_of edge to their parent enum, covering explicit and implicit values, const enum, and quoted TypeScript member names — consistent with existing Java/Kotlin/Swift enum handling.

Installing graphifyy[postgres] now includes the tree-sitter-sql grammar required by the SQL introspection path, and a missing or ABI-incompatible grammar raises an actionable error instead of silently returning zero nodes.

Multi-repo community detection now offsets each input's community ids so community 0 of one repo no longer fuses with community 0 of another; the original id is preserved in the local_community field.

Performance: ignore-pattern evaluation in pattern-heavy monorepos is dramatically faster — relative paths are now computed lexically, each pattern is parsed once into a bounded process cache, and per-entry work is memoized; ignore decisions are unchanged.

Running graphify extract --code-only --force over an existing graph now preserves the document/paper/image semantic layer instead of dropping it; files deleted from disk are still pruned.

graphify merge-graphs now links a type declaration shared across two repos — matched by fully-qualified namespace and name — with a same_type_as edge, making shared contract types navigable across repo boundaries.

C# object-creation expressions (new Foo()) now emit an edge to the constructed type, making constructor usage visible in the graph; built-in and out-of-corpus types are not fabricated.

C# interfaces with exactly one implementing class now have their methods linked to that implementation with a dispatches_to edge, so calls through injected dependencies reach the concrete implementation.

Supports binding app.message.followUp to alt+enter in pi config to use Alt+Enter for follow-up queueing instead of Ctrl+Q.

Adds a powershell tool that runs commands through pwsh.exe (falling back to Windows PowerShell) with -NoProfile -NonInteractive -ExecutionPolicy Bypass; administrator-enforced execution policies still take precedence.

Supports configuring defaultTools to replace or supplement the default bash tool with powershell on Windows — e.g. "defaultTools": ["read", "powershell", "edit", "write"] to replace, or "defaultTools": ["read", "bash", "powershell", "edit", "write"] to run both side by side.

Adds /thinking selector with searchable default choices to the model and thinking selectors; Ctrl+S saves the selected model as the global default, persisting it session-wide.

Adds variant config key to set provider-specific reasoning effort (e.g. high, max, or minimal) for the GitHub Action.

Adds oidc_base_url config key to point the OIDC token exchange at a custom GitHub App installation instead of the default https://api.opencode.ai.

Adds mentions config key accepting a comma-separated, case-insensitive list of trigger phrases (defaults to /opencode,/oc) to control which PR/issue comments activate the bot.

Supports caller-provided token mode that skips OIDC exchange entirely, allowing a PAT or third-party GitHub App token stored as a secret to be passed directly — removing the requirement for id-token: write permission.

New gemini skills install <url|path> --consent CLI command installs a skill from a Git repository or local directory, with --scope user|workspace to control whether it is global or project-local.

New gemini skills list --all CLI command lists all discovered skills across every tier, including built-in skills normally hidden from output.

New /skills list [all] [nodesc] interactive session command shows discovered skills, with all to surface built-in skills and nodesc to suppress descriptions.

New /skills link <path> [--scope user|workspace] interactive session command links skills from an arbitrary local directory into the active session.

New /skills disable <name> and /skills enable <name> interactive session commands toggle individual skills without uninstalling them.

New /skills reload (alias /skills refresh) interactive session command rescans all discovery tiers and refreshes the skill list mid-session.

Agent Skills are discovered from four tiers in precedence order: built-in, extension, user (~/.gemini/skills/ or ~/.agents/skills/), and workspace (.gemini/skills/ or .agents/skills/), with workspace skills version-controllable for team sharing.

Adds the activate_skill tool — called automatically by the model when a task matches a skill's description — which injects the skill's SKILL.md and grants the agent read access to the skill's bundled assets.

Skills support an interoperable .agents/skills/ path alias so skill directories remain compatible across multiple AI tools beyond Gemini CLI.

Adds media_storage parameter (type Optional[Union[MediaStorage, AsyncMediaStorage]]) to Agent, enabling offloading of media to external storage while keeping only a reference in the database.

Adds delete_media parameter (bool, default False) to Agent, which when True also deletes a session's offloaded media from media_storage on session deletion.

New Media Storage subsystem for persisting session media, with backends for local filesystem, Amazon S3 (S3 Media Storage), and Google Cloud Storage (GCS Media Storage).

Adds support for registering existing or vendor-supplied MCP servers with Arcade, enabling unified governance of external MCP tools alongside native Arcade tools.

Adds qwen/qwen3.8-27b (Qwen 3.8 27B) as a supported vision model — a 27B multimodal model accepting up to 3 images per request, with thinking and instruct modes, tunable reasoning effort, tool use, and JSON mode; each image counts as 2048 input tokens.

Adds support for Qwen3 models, with qwen/qwen3.8-27b supporting low, medium, and high reasoning effort levels; high maps to the model's native xhigh mode.

New POST /v2/{endpoint_id}/batch endpoint creates a batch job (optionally with initial requests) that runs asynchronously on dedicated workers, isolated from standard /run traffic.

New POST /v2/{endpoint_id}/batch/{id}/requests endpoint appends additional requests to a DRAFT batch incrementally; each call accepts up to 10 MiB of request body.

New GET /v2/{endpoint_id}/batch/{id} endpoint returns batch summary with progress fields requestTotal, requestInProgress, requestCompleted, and requestFailed.

New GET /v2/{endpoint_id}/batch/{id}/requests endpoint returns paginated child request results; supports offset and limit query parameters with a hasMore field for pagination.

Batch lifecycle moves through states DRAFT → FINALIZED → FAILED / CANCELLED; completion is inferred when requestCompleted + requestFailed equals requestTotal.

Console Inbox notifications are sent on terminal state transitions, including batch ID, endpoint name, status, and item counts.

New batch jobs capability for RunPod Serverless allows submitting large collections of inference requests as a single named batch for asynchronous processing.

Bring Your Own Cloud (BYOC) is now generally available on the Enterprise plan across AWS, GCP, and Azure — runs the Pinecone data plane inside your own cloud account so vectors, metadata, and queries never leave your environment, with an in-cluster agent pulling operations from Pinecone without requiring inbound network access.

Pinecone Nexus, a knowledge engine for agents, is now generally available — point it at your sources, curate them into a context, and query that context to receive grounded, cited answers.

New upsert_file API endpoint (2026-04): creates or replaces a file in an assistant by assistant_file_id, enabling idempotent file uploads.

New list_operations API endpoint (2026-04): lists all in-progress, completed, and failed ingestion operations for an assistant, filterable by operation_type and status query parameters; operations are retained for 30 days.

New metrics_alignment API endpoint evaluates RAG answer correctness, completeness, and alignment (harmonic mean of correctness and completeness) against a ground-truth answer.

Adds an OpenAI-compatible Chat Completion API interface for Pinecone Assistant, enabling inline citations, streaming responses, and drop-in integration with OpenAI-compatible clients.

Adds a Model Context Protocol (MCP) server for Pinecone Assistant, supporting both remote and local MCP endpoints for connecting AI agents (including Cursor and Claude Desktop).

Adds context snippet retrieval from Pinecone Assistant to power external LLMs, RAG pipelines, or agentic workflows, returning citations with signed URLs.

Adds an n8n workflow quickstart for Pinecone Assistant integrating with OpenAI to download files via HTTP, upload documents, and chat with them from an automation.

Supports spend alerts and per-dimension usage monitoring (chat tokens, context retrieval, evaluation) for Pinecone Assistant organizations.

New storage integrations enable bulk import of vector data into indexes and export of audit logs from Amazon S3 (via IAM role), Google Cloud Storage (via service account key), and Azure Blob Storage (via service principal).

New Pinecone Nexus product supports BYOC (Bring Your Own Compute) deployment, KnowQL queries that return grounded multi-document answers with citations, and per-query tracing showing reasoning steps, retrieval tool calls, token usage, latency, cost breakdown, and cache savings.

New Standard plan trial provides $300 in credits over 21 days, including bulk import, backup and restore, RBAC, and higher scale limits.

Adds DefaultBraintrustApiClient.Api, a generated client covering every Braintrust REST endpoint, shipped inside the existing Braintrust.Sdk package with no separate install required.

Web search, file search, code interpreter, and MCP calls in the OpenAI Responses API now appear as child tool spans nested under the LLM span for cleaner trace visualization.

server_tool_use blocks in the Anthropic Messages API now appear as child tool spans under the LLM span, and Anthropic spans capture new server_tool_use_* metrics.

Hovering a row in the List view of a Topics facet card now reveals a 'View logs' link that filters logs to that topic and all nested topics.

Custom preprocessors whose output is not a recognized message array now render in thread layout as labeled text instead of failing.

New Topics feature automatically classifies production logs by user intent, sentiment, and issues — with built-in Task, Sentiment, and Issues facets — then surfaces patterns as topic distributions and time-series trends.

Custom facets let teams define domain-specific classification dimensions using custom preprocessors and prompts beyond the built-in Topics facets.

Adds dataset pipelines to transform spans or traces from project logs into dataset rows in bulk using declarative, version-controlled pipelines.

Remote evals and sandboxes allow connecting custom agent code to a playground and running it in an isolated sandbox to test complex agents interactively without modifying deployed code.

Supports evaluating reasoning models (OpenAI o4, Claude, Gemini) with configurable thinking budgets, capture of intermediate reasoning traces, and scoring of final outputs.

New Loop agent enables natural-language interaction to analyze logs, optimize prompts, build datasets, and generate custom React view components.

Custom views let teams describe a trace/dataset interface in natural language; Loop generates an embeddable, customizable React component.

Advanced tracing patterns now cover distributed tracing across services, sensitive data masking, custom span rendering, and high-throughput performance tuning.

Log attachments support images, audio, PDFs, and other binary files attached to traces and experiments for multimodal review in context.

Topics management includes on-demand pipeline runs, sampling adjustment, pause/resume, and history rewind after facet or filter changes.

Human review now supports multiple reviewers with configurable score-combination rules and a base value for non-numeric scores.

Dataset performance tracking shows which experiments used a dataset and how each row performs over time, with chart and SQL filters.

Adds export of annotated traces and datasets for use in external evaluation frameworks, analysis pipelines, or training workflows.

The SQL sandbox now organizes queries in a collapsible sidebar with search by name, drag-to-reorder, command-bar navigation, and a per-query menu to rename, duplicate, or delete.

Manage a member's permission groups directly from the member dialog — showing direct groups, inherited groups, and available groups to add — without opening each group individually.

Filter the logs list to a specific span name by clicking that span's name in the trace panel header, instead of building the filter manually.

Adds AWS IAM role assumption support for S3 bulk-export destinations on GCP-hosted LangSmith SaaS, allowing credential-free exports using a trusted AWS role ARN instead of static AWS credentials.

Supports switching an existing S3 export destination between static credentials and IAM role assumption without recreating the destination.

New Chat interface on the Engine page lets practitioners ask natural-language questions about detected issues and their linked traces, such as 'What went wrong?', 'Why was this flagged?', 'Explain the proposed fix', 'What are my most pressing issues?', and 'How many new issues do I have?'

Adds aws_role_arn field to the PATCH body for S3 export destinations, enabling LangSmith to assume an AWS IAM role instead of storing static credentials — available on GCP SaaS deployments only.

Switching authentication modes (static credentials ↔ IAM role assumption) is now possible without recreating the destination: supplying aws_role_arn clears stored credentials, and supplying static credentials clears the stored role ARN.

LangSmith validates any authentication-mode switch with a test write to the destination bucket before saving, returning an error if the new configuration lacks sufficient write permissions.

Exports AI-BOM data in industry-standard CycloneDX 1.7 and SPDX 3.0.1 formats, with metadata-only exports to preserve data privacy.

Introduces Account-Level (Global) Workflows, letting administrators define workflows centrally and propagate them to selected organizations, with controls over whether org-level admins can enable or disable them locally.

Updates the Administration page sidebar with a categorized, collapsible navigation structure, adding a dedicated 'AI Usage' tab under a new AI category and sub-headers for AI, Code, and Dependencies in Engine Settings.

Adds Chainguard as a supported vendor in Mend Container's hardened image detection, incorporating Chainguard-specific vulnerability data, VEX information, and detection of Chainguard image layers and base images.

Surfaces Chainguard VEX status and branding directly in the risk factors columns across the UI for Chainguard-based images and findings.