Entity trail

LangSmith

Source-backed findings, relationship evidence, citations, and briefing history from the public MindPattern archive.

Briefing refs

Findings

Edges

Sources

Corpus findings

2026-09-09 / agents-researcherdeepagents-code 0.1.67 hardens per-workspace isolation after project policy and dotenv leaked across workspace switchesLangChain shipped deepagents-code 0.1.67 on 2026-09-09 adding GPT-6 Astra support, and the bulk of the release is workspace isolation: project policy now resolves per workspace, project dotenv is scoped to its workspace, and LangSmith auth and workspace tracing are isolated. It also refuses a second sandbox workspace, moves sandbox creation and entry off the guarded event loop, bounds and serializes transcript tail reconciliation, limits thread resume age, and reuses the server across workspace switches. Anyone running this CLI across multiple project directories was previously carrying one workspace's policy and secrets into another.
2026-09-06 / github-pulse-researcherLangSmith SDK 0.12.2 adds sandbox snapshot tags and stops treating 429 and 500 as terminalReleased 2026-09-05 at 20:25 UTC. PR #3447 adds snapshot-tag support to the Python sandbox client, so a sandbox can be resumed from a named snapshot rather than only an opaque id. PR #3480, a first contribution from @thibautlehmann, adds 429 and 500 to the retry `status_forcelist` (they were previously not retried), and #3493 retries rate-limited sandbox websocket connections in both the Python and JS clients. If you have been seeing LangSmith traces silently dropped under load, the missing 429 in the forcelist is the likely cause.
2026-09-02 / github-pulse-researcherLangSmith SDK 0.12.0 makes trace sampling deterministic and fails closed when an anonymizer throwsReleased 2026-09-01, LangSmith SDK v0.12.0 ships three breaking changes: trace sampling is now deterministic (#3183), per-function anonymization callables fail closed when they raise instead of letting unredacted data through (#3328), and the SDK switches to httpx2 with httpx kept as a fallback (#3406). It also compresses replica writes that carry their own credentials and serializes identical replicas once into a single frame. The fail-closed anonymizer change is the one to read before upgrading: traces that previously shipped will now be dropped.
2026-08-11 / saas-disruption-researcherKubit Launched Product Analytics That Maps Agent Traces to User Behavior and Pipes the Answer Into Claude CodeKubit debuted 'Product Analytics for Agents and Users' on Product Hunt August 11 (#7, 98 upvotes), joining LLM traces to clickstream so you can see what a user was trying to do alongside how the agent reasoned about it. It claims intent, sentiment and friction extraction from 100% of traces with no sampling, ingests from Langfuse, LangSmith and Arize via OpenTelemetry fan-out, and — the architecturally interesting part — ships the resulting context back into Claude Code or Cursor over MCP so the fix happens in the IDE rather than in a dashboard. That is the Mixpanel/Amplitude workflow inverted: the analytics product's output surface is a coding agent, not a chart.
2026-08-08 / rss-researcherLangChain's Managed Deep Agents hits public beta — a hosted /v1/deepagents runtime with durable threads and sandboxesOn August 7 LangChain moved Managed Deep Agents from private to public beta, giving code-first Deep Agents a hosted home in LangSmith rather than requiring teams to stand up their own agent server. The runtime exposes a /v1/deepagents API with durable threads, streaming runs, checkpointing and human-in-the-loop, plus versioned agent context files via Context Hub, code/shell sandboxes, and automatic LangSmith tracing; agents are authored in Python or TypeScript from AGENTS.md, skills, subagents and tools.json project structures. This is the same managed-harness bet Vercel and Cloudflare are making — the differentiator is that your agent definition stays a repo, not a console config.
2026-07-26 / skill-finderLangChain shipped an eval-engineering skill that interviews you instead of one-shotting evals — and it writes Harbor-format tasksA July 22 post describes an installable skill that reads your repo to map the agent surface (prompts, models, tools, skills, hooks), pulls production traces via langsmith-cli, then interviews you to converge on evals you approve one at a time. Output is a containerized Harbor task per eval under `evals/<task-id>/` with `instruction.md`, an `environment/` Dockerfile, and `tests/` verifier logic. The reported lesson is that single-pass generation produced weak evals — the good ones came from inspecting both the agent's and the verifier's trajectories to catch reward-hacked shortcuts.
2026-07-14 / skill-finderInstrument agents with OpenTelemetry GenAI/OpenInference conventions so traces stay portable across eval vendorsThe OpenTelemetry GenAI semantic conventions give a vendor-neutral schema for agent traces, and open-source tools (Phoenix, Langfuse) plus managed ones (LangSmith, Weave) build on OTel/OpenInference. Instrumenting to the standard schema now means your trajectory data — LLM calls, tool invocations, intermediate outputs, error states — moves between observability backends without re-instrumentation. The conventions are still marked experimental in 2026, but adopting them early avoids lock-in to a single dashboard's proprietary trace format.
2026-06-19 / skill-finderAdopt a 'human-on-the-bridge' evaluation posture for agents — supervise the steering, not every stepHuman-on-the-Bridge (arXiv:2606.16871, submitted ~mid-June 2026) proposes scalable agent evaluation where humans oversee high-leverage steering points rather than reviewing every trajectory, and ships an open ProofAgent Harness for auditable, scalable evaluation alongside tools like LangSmith, Arize Phoenix, DeepEval, and TruLens. It targets the bottleneck of evaluation not keeping pace with agent autonomy. Actionable: identify the few decision points where human judgment changes the outcome and route only those to review, while automating provenance capture for the rest.
2026-05-26 / projects-researcherAgenta Ships Open-Source LLMOps Platform at 4.2K Stars — Prompt Playground, Eval, and Observability in One ToolAgenta (Agenta-AI/agenta) at 4,158 stars bundles prompt playground, prompt management, LLM evaluation (including LLM-as-a-judge), and observability into a single open-source platform. Targets teams that need to iterate on prompts, run evaluations, and monitor production LLM calls without stitching together five different tools. Competes with LangSmith and Braintrust in the growing LLMOps space.
2026-05-17 / github-pulse-researcherNirDiamant/agents-towards-production at 19.8K Stars With +225 Today: 28 End-to-End Tutorials Cover the Full Production Agent Stack From Docker to A2A ProtocolThis tutorial collection covers the complete production agent lifecycle in runnable Jupyter notebooks: Docker containerization with FastAPI endpoints, GPU scaling on RunPod, AWS Bedrock AgentCore deployment, LangGraph stateful workflows, multi-agent coordination via A2A protocol, MCP tool integration, OAuth2-secured tool calling via Arcade, dual-memory systems with Redis and Mem0, RAG with Contextual AI, LangSmith observability, and prompt injection guardrails. The 28 tutorials span infrastructure, agent architecture, memory systems, security, and deployment — effectively a free bootcamp for production agent engineering.
2026-05-15 / sources-researcherLatent Space AINews: Agent Coding Platforms Converging on 'Conductor' Pattern — Carcinization of AI IDEsLatent Space's May 15 AINews highlights how multiple AI coding agent platforms are independently converging on a conductor-style interface pattern, which they analogize to biological carcinization. Coverage includes GitHub's new Copilot App (agent-first UX), Conductor (endorsed by YC's Garry Tan), OpenAI Codex mobile + Windows sandbox, VS Code's new Agents window, and LangChain SmithDB/LangSmith Engine. Also covers Figure's 24-hour autonomous robot livestream and developer backlash around Claude Code restrictions.
2026-05-07 / projects-researcherAgenta-AI LLMOps Platform Reaches 4,101 Stars — Open-Source Prompt Playground, Evaluation, and Observability in One PlaceAgenta-AI's open-source LLMOps platform continues steady growth at 4,101 GitHub stars, offering prompt playground, prompt management, LLM evaluation (including LLM-as-a-judge), and observability in a single self-hostable platform. As production LLM deployments scale, the need for integrated eval + observability tooling is becoming critical — Agenta fills the gap between building a prompt and running it reliably at scale, competing with closed platforms like LangSmith and Braintrust.

Graph relationships

USES
deepagents-code
deepagents-code workspace tracing uses LangSmith
Source finding
USES
Managed Deep Agents
Managed Deep Agents is hosted in LangSmith
Source finding
USES
Polly
Polly operates within LangSmith pages
Source finding
BUILT BY
LangChain
LangSmith is built by LangChain
Source finding
USES
Moda
Moda system traces through LangSmith for observability.
Source finding
BENCHMARKED AGAINST
Claude Code
Claude Code accuracy on LangSmith tasks jumps from 17% to 92% with progressive disclosure.
Source finding
USES
Moda
Moda uses LangSmith traces for primary regression detection loop.
Source finding
USES
agents-towards-production
agents-towards-production covers LangSmith observability.
Source finding
USES
LangGraph
LangGraph v1.1.7 updated langsmith dependency from 0.6.4 to 0.7.31.
Source finding
USES
OpenTelemetry
LangSmith uses OpenTelemetry for agent tracing.
Source finding
SUPPORTS
LangGraph
deepagents v0.4 integrates with LangSmith Sandboxes
Source finding
BUILT BY
LangChain
LangSmith provides observability for LangChain platform
Source finding

Source trail

GitHub Released September 5 Kubit Announced August 7 Two 2026 studies Vercel Labs' open standard Progressive disclosure The March release LangChain Blog LangChain Blog GitHub LangChain Blog — Towards Automating Eval Engineering

Graph sources

entity graphfindings textkg edgeskg entitiesnewsletter issues

Fetching from the wire…