Observed Signal · Jul 3, 2026 · Technical Publication · Source: DEV Community · Impact: 2/5 · Sentiment: Positive
ReWOO: Plan Tool Calls Upfront, Only Two Model Calls
The article introduces ReWOO (Reasoning WithOut Observation), an agent design that asks an LLM to plan every tool call up front, executes the planned tool calls in plain code without further model calls, then calls the LLM once more to produce the final answer. By splitting the agent into a Planner (one LLM call), a Worker (zero LLM calls) and a Solver (one LLM call), ReWOO keeps model-call count constant at two regardless of chain length, reducing token retransmission, inference cost and latency for multi-hop queries. The piece explains the trade-off: ReWOO is efficient when the step sequence is predictable but less adaptable to surprising or errorful tool results, and it suggests hybrid approaches (re-planning on surprises). Published on dev.to on 2026-07-03.
Presents a practical agent architecture that materially reduces LLM calls, tokens and latency for multi-hop workflows — relevant for teams deploying cost-sensitive RAG/agent systems — but is not a platform-level or industry-shifting announcement.
Track DEV Community Signals & Market Shifts in Real-Time
Polaris7 autonomous intelligence agents track regulatory filings, primary sources, executive changes, and deal flow 24/7. Create your free Explorer workspace to monitor these entities.
Key Takeaways & Evidence Grounding
- ReWOO (Reasoning WithOut Observation) asks an LLM to produce a complete plan of tool calls up front and then performs the tool calls in code before a final LLM call.
- ReWOO splits the agent into three roles: Planner (1 LLM call), Worker (0 LLM calls), and Solver (1 LLM call).
- Model-call count under ReWOO is constant at 2 regardless of the number of hops; ReAct typically requires k+1 model calls for k hops.
- ReWOO reduces repeated transcript token usage and can lower latency and inference cost, but it sacrifices mid-run adaptability if tool results are unexpected.
- Article published on dev.to with a metadata publication date of 2026-07-03.
Connected Companies & Entities
1 Entity mappedOntology Mapping & Concepts
Related Market Signals & Shifts
Recent verified developments and strategic activity across this market segment.
Plan-and-Solve Agent Architecture: Plan First, Then Execute
This technical article explains the Plan-and-Solve agent paradigm: use an LLM to generate a complete, ordered plan of 3–7 concrete steps (Plan phase), then execute each step sequentially with tool-enabled sub-agents (Solve phase). The author implements the architecture using LangGraph's StateGraph abstraction, showing a Plan node, an Execute node that embeds a ReAct sub-agent (for tool calls like web_search and calculator), a Replan node, and a Finalize node. Demos highlight practical failure modes—information loss when step results are transmitted as short natural-language summaries, planner over-splitting, and tool-specific errors—and propose engineering fixes (structured step outputs, dedicated collected_data state, planner-annotated data flow). The article concludes with five production findings and guidance on when to choose ReAct versus Plan-and-Solve.
Open‑Source RightModel Tackles Token Consumption Anxiety
Developer Regnard Raquedan published an article on May 4, 2026 describing RightModel, an open-source tool that recommends the best LLM for a task without making live LLM calls in the default request path. RightModel uses a human-owned, versioned ruleset to classify task types and map them to model tiers; pricing data is refreshed asynchronously (via OpenRouter and a scheduled workflow using Google Cloud Scheduler) to avoid runtime API calls. For ambiguous cases the app exposes a user-triggered "Deep Analysis" escalation that calls an LLM (currently Gemini 2.5 Flash). Raquedan frames the architecture as an instance of a broader pattern he calls "Precomputed AI," which shifts reasoning out of real-time request paths into asynchronous build pipelines with explicit staleness controls and escalation paths.
ARTIST: RL-Powered Tool Use for LLM Agents
ARTIST (Agentic Reasoning and Tool Integration in Self-improving Transformers) is a Microsoft Research training framework that teaches LLMs when and how to call external tools by using outcome-only reinforcement learning. Published (paper/arXiv 2505.01441) and described in the article, ARTIST interleaves tool calls inside the model’s chain-of-thought tokens rather than appending results as separate turns, and trains with GRPO (Group Relative Policy Optimization) using final-answer rewards, format checks, and efficiency signals. At 7B scale, ARTIST reportedly outperforms GPT-4o on multiple math and multi-turn function-calling benchmarks. An independent Effloow Lab proof-of-concept reproduced the interleaving execution loop in a minimal Python sandbox and observed improved accuracy and fault-tolerant recovery. The paper focuses on a training recipe rather than a production SDK; some public implementations (TRL, verl) can approximate parts of the GRPO loop.
Track Real-Time Market Signals & Shifts
Set up custom watchlists to receive automated, evidence-grounded executive digests whenever material signals or shifts occur across your tracked landscape.