GraphDoc
We have open sourced graphdoc
under the Apache 2.0 License
, and the repository can be found here. Some things we want to highlight about graphdoc
is that it: 1) serves as a working production implementation of a self-improving, prompt-based LLM workflow for the community, 2) includes a containerized version of MLFlow
, following industry best practices, so that anyone can locally or remotely run, track, and reproduce prompt optimization experiments, and 3) is accompanied by a production ready server, running an instance of graphdoc
, that is made freely available to the public (link). We have designed this system with continuous improvement in mind and as we receive documentation requests, we plan to continue to refine our documentation pipeline accordingly. There is an open discussion in graphprotocol/graph-tooling regarding the integration of graphdoc
into the graph-cli
.
GraphTally and Indexer Service
Throughout the past month, the team has focused on preparing GraphTally and indexer service for the upcoming Horizon upgrade. We began with deep-dive sessions—first internally, then with protocol engineers at Edge & Node—to align on the upgrade’s implications and plan a test-driven path forward.
A major area of focus has been understanding the current indexer-agent
component, a TypeScript service that remains critical to how GraphTally operates. We’ve been mapping out how it fits into Horizon-era workflows and what will be required to support and eventually evolve it.
At the same time, we’ve invested heavily in improving our testing infrastructure. We’ve implemented a local testnet that is already proving its value: we used it to reproduce and fix a user-reported crash, and it’s now helping us investigate further issues. We also began integrating profiling support to improve observability. To support broader testing goals across the stack, we met with Edge & Node’s gateway and networking teams to share our approach and align on how the local-network testing tool can be used across components. This work is not only helping us prepare for Horizon, but also directly improving our ability to maintain the current network. As we expand test coverage and sharpen our visibility into performance, we’re identifying potential issues earlier and improving the overall stability of GraphTally and the indexer service.
We also gave an update and answered questions at Indexer Office Hours, helping share context on the work underway and how it supports the protocol’s upcoming changes. We showed some successful examples of recent interactions with Indexers, and by doing so, we hope to tighten the feedback loop between our work and the experience of Indexers on the network.
Looking ahead, we’ll continue to refine our test tooling and work closely with protocol teams to ensure a smooth transition as Horizon smart contracts are finalized.
Verifiable ELTQ
Verifiable Extraction
We have updated veemon
with libraries that can be used to verify Arbitrum and Solana extracted data. We are currently cleaning up the library, documenting and writing examples to demonstrate how this library can be integrated in future services provided by The Graph.
Verifiable Querying
We have stopped further development in this area.